From dina.papagiannaki at intel.com Fri Apr 11 09:54:10 2008 From: dina.papagiannaki at intel.com (Papagiannaki, Dina) Date: Fri, 11 Apr 2008 17:54:10 +0100 Subject: [e2e] CFP: Internet Measurement Conference (IMC) 2008 Message-ID: <708F8794D464F346B28DF53E5CDD805F038A5462@swsmsx412.ger.corp.intel.com> Please note that the registration deadline is fast approaching - May 2nd 2008. ======================================================================== ====== Internet Measurement Conference (IMC) 2008 Sponsored by ACM SIGCOMM and in cooperation with USENIX October 20-22, 2008 Vouliagmeni, Greece The Eighth Internet Measurement Conference is a two and a half day event focusing on Internet measurement and analysis, building on the success of past IMCs. We invite submissions of papers that contribute to our understanding of how to collect or analyze Internet measurements, or give insight into Internet structure and behavior. Examples of relevant topics are: * Internet traffic analysis * Internet structure and topology characteristics * Internet performance measurements * Measurement-based network management such as traffic engineering * Inter-domain and intra-domain routing * Network applications such as WWW, multimedia streaming, and gaming * Measurements of content distribution, peer-to-peer, overlay, and social networks * Data-centric issues, including anonymization, querying, and storage * Measurement-based inference of network properties * Design of monitoring systems, sampling methods, signal processing methods * Network anomaly detection * Network security threats and countermeasures * Software tools and environments in support of measurement * Measurement-based assessment of simulation/testbeds * Measurement-based workload generation * Measurement-based modeling * Reappraisal of previous measurement findings * Internet-oriented wireless, and mobility measurement Papers that do not in some fashion relate to measuring Internet properties are out of scope. Authors can contact the Program Chair at imc08chairs at cs.umn.edu for clarification if they are unsure whether their paper is in scope. Important dates * May 2, 2008: 10PM EDT: Registration of title and 250-word abstract * May 9, 2008: 10PM EDT HARD submission deadline * July 18, 2008: Notification * August 27, 2008: Camera Ready Copy due * October 20-22, 2008: Conference held in Vouliagmeni, Greece For more details please see http://www.imconf.net/imc-2008/cfp.html From Jon.Crowcroft at cl.cam.ac.uk Fri Apr 18 03:10:26 2008 From: Jon.Crowcroft at cl.cam.ac.uk (Jon Crowcroft) Date: Fri, 18 Apr 2008 11:10:26 +0100 Subject: [e2e] end of interest Message-ID: last month's postings on this list was 80% CfPs; this month has 1 message which is a CfP; is this the end of interest in end-to-end? has the location/identifier debate mean everyone split? does virtualization mean unjustifying the ends? has mesh networking for nanotech or packet swarms and multipath routing diverted attention elsewhere? its not like there's no work out there - just look at this weeks NSDI programme:- http://www.usenix.org/event/nsdi08/tech/ j. From dpreed at reed.com Fri Apr 18 06:13:57 2008 From: dpreed at reed.com (David P. Reed) Date: Fri, 18 Apr 2008 09:13:57 -0400 Subject: [e2e] end of interest In-Reply-To: References: Message-ID: <48089E95.1090103@reed.com> I personally think that the network community has become frustrated with the inability to explore end-to-end protocols because the endpoint stacks are "locked in" by vendors in proprietary code. Jon Crowcroft wrote: > last month's postings on this list was 80% CfPs; > this month has 1 message which is a CfP; > is this the end of interest in end-to-end? > > has the location/identifier debate > mean everyone split? > > does virtualization mean unjustifying the ends? > > has mesh networking for nanotech > or packet swarms and multipath routing > diverted attention elsewhere? > > its not like there's no work out there - just look at > this weeks NSDI programme:- > http://www.usenix.org/event/nsdi08/tech/ > > j. > > From rsofia at inescporto.pt Fri Apr 18 08:26:21 2008 From: rsofia at inescporto.pt (Rute Sofia) Date: Fri, 18 Apr 2008 16:26:21 +0100 Subject: [e2e] end of interest In-Reply-To: References: Message-ID: <4808BD9D.2070800@inescporto.pt> Jon, (silence is also music :D) Some interesting papers, even though I'm specifically referring to the one from Dina Katabi, FatVAP. Clever and simple user-centric idea to perform load-balancing among neighboring APs. Rute Jon Crowcroft wrote: > last month's postings on this list was 80% CfPs; > this month has 1 message which is a CfP; > is this the end of interest in end-to-end? > > has the location/identifier debate > mean everyone split? > > does virtualization mean unjustifying the ends? > > has mesh networking for nanotech > or packet swarms and multipath routing > diverted attention elsewhere? > > its not like there's no work out there - just look at > this weeks NSDI programme:- > http://www.usenix.org/event/nsdi08/tech/ > > j. From dpreed at reed.com Fri Apr 18 13:47:08 2008 From: dpreed at reed.com (David P. Reed) Date: Fri, 18 Apr 2008 16:47:08 -0400 Subject: [e2e] end of interest In-Reply-To: References: <48089E95.1090103@reed.com> Message-ID: <480908CC.5050400@reed.com> Andrew - I do exploration in Linux myself (I only run Windows in KVM, and only when I absolutely have to). Regarding the Major Company's proprietary TCP code - I agree that minor tweaks to the TCP standard are adopted as long as the Major Company's programmers are the ones who act as gatekeeprs. However, that's like what AT&T Bell Labs used to say when they disparaged the Internet Experiment in the 1980's - "when those lunatics working with Vint Cerf and Bob Kahn do something really interesting, we'll put it in our network". And of course we have people on this list who view any attempt to deploy on the public Internet experiments among early adopters of Linux kernels (like the Fedora distribution) as "evil things" to be stopped as a matter of policy. All of this puts grit in the gears of innovation by requiring that innovators "get permission" from those who want to preserve their market power. Sometimes there is an actual risk, but mostly it's fearmongering like the AT&T argument against Hush-a-Phone. (a plastic piece you added to a Bell phone that was reputed to have the potential of damaging the network). If you have no hope of deploying most innovations without bargaining with Major Co., then why bother doing research? - David Lachlan Andrew wrote: > Greetings David, > > Exploration can be done in Linux. Talking to an author of a Major > Company's proprietary TCP code, he seemed quite open to implementing > IETF standards, provided they work with the middle boxes. > > Jon, > > One interesting end-to-end issue is congestion control. That > discussion has moved to the iccrg mailing list, which is quite > active. > > Cheers, > Lachlan > > On 18/04/2008, David P. Reed wrote: > >> I personally think that the network community has become frustrated with the >> inability to explore end-to-end protocols because the endpoint stacks are >> "locked in" by vendors in proprietary code. >> >> >> >> Jon Crowcroft wrote: >> >> >>> last month's postings on this list was 80% CfPs; >>> this month has 1 message which is a CfP; >>> is this the end of interest in end-to-end? >>> >>> has the location/identifier debate mean everyone split? >>> >>> does virtualization mean unjustifying the ends? >>> >>> has mesh networking for nanotech >>> or packet swarms and multipath routing diverted attention elsewhere? >>> >>> its not like there's no work out there - just look at >>> this weeks NSDI programme:- >>> http://www.usenix.org/event/nsdi08/tech/ >>> >>> j. >>> >>> >>> >>> > > > From saikat at cs.cornell.edu Fri Apr 18 15:13:21 2008 From: saikat at cs.cornell.edu (Saikat Guha) Date: Fri, 18 Apr 2008 18:13:21 -0400 Subject: [e2e] end of interest In-Reply-To: <480908CC.5050400@reed.com> References: <48089E95.1090103@reed.com> <480908CC.5050400@reed.com> Message-ID: <1208556801.23430.2.camel@sioux.systems.cs.cornell.edu> On Fri, 2008-04-18 at 16:47 -0400, David P. Reed wrote: > If you have no hope of deploying most innovations without bargaining > with Major Co., then why bother doing research? There are multiple major companies with competing interests ?though. Research that favors one would get deployed, or at least, would have a chance of getting deployed. Google and network neutrality for example. -- Saikat -------------- next part -------------- A non-text attachment was scrubbed... Name: not available Type: application/pgp-signature Size: 197 bytes Desc: This is a digitally signed message part Url : http://mailman.postel.org/pipermail/end2end-interest/attachments/20080418/24a123f4/attachment.bin From mahesh at cs.cornell.edu Fri Apr 18 17:11:29 2008 From: mahesh at cs.cornell.edu (Mahesh Balakrishnan) Date: Fri, 18 Apr 2008 20:11:29 -0400 Subject: [e2e] end of interest References: <48089E95.1090103@reed.com> Message-ID: In fact there seems to be pushback from both ends --- we can't deploy end-to-end protocols because major companies own the end-host stacks; and we can't push mechanisms deep into the network because ISPs and router companies own the network. Arguably the latter source of pushback played a major role in the emergence of the e2e philosophy; but now we have equally powerful commercial forces on the other side. So effectively the only practical mode of deployment seems to be the 'almost' end-to-end middlebox --- one hop away from the end-host but not quite into the network (and the Maelstrom work I presented day before yesterday at NSDI would be one example). - mahesh -- http://www.cs.cornell.edu/~mahesh -----Original Message----- From: end2end-interest-bounces at postel.org on behalf of David P. Reed Sent: Fri 4/18/2008 9:13 AM To: Jon Crowcroft Cc: 'end2end-interest at postel.org' Subject: Re: [e2e] end of interest I personally think that the network community has become frustrated with the inability to explore end-to-end protocols because the endpoint stacks are "locked in" by vendors in proprietary code. -------------- next part -------------- An HTML attachment was scrubbed... URL: http://mailman.postel.org/pipermail/end2end-interest/attachments/20080418/fb169fc0/attachment.html From moore at cs.utk.edu Fri Apr 18 19:40:36 2008 From: moore at cs.utk.edu (Keith Moore) Date: Fri, 18 Apr 2008 22:40:36 -0400 Subject: [e2e] end of interest In-Reply-To: References: <48089E95.1090103@reed.com> Message-ID: <48095BA4.3020709@cs.utk.edu> Seems to me like the biggest barrier to deploying new L3/4 end-to-end functionality isn't host stacks, but rather middleboxes that don't know what to do with new IP packet types or protocol extensions. If you can define a new e2e feature in such a way that it traverses those, and which doesn't need explicit support from the existing IP network, it's still possible to deploy it. Granted you won't see widespread use of the new feature unless the 900 pound gorilla ships it in product, but Linux is used widely enough that it's possible to use it as a vehicle to prove that a feature is valuable. If you could come up with an enhancement that, say, improved HTTP performance and you could get it in the Linux kernel (which is much easier than convincing the gorilla to ship it), the gorilla would feel the need to follow suit or improve on that enhancement -- to avoid embarrassing itself further. Though I do think there's a place for the 'almost' e2e middlebox, particularly in implementing layered defense against security threats and wasted bandwidth from unwanted traffic. Keith > > In fact there seems to be pushback from both ends --- we can't deploy > end-to-end protocols because major companies own the end-host stacks; > and we can't push mechanisms deep into the network because ISPs and > router companies own the network. Arguably the latter source of > pushback played a major role in the emergence of the e2e philosophy; > but now we have equally powerful commercial forces on the other side. > > So effectively the only practical mode of deployment seems to be the > 'almost' end-to-end middlebox --- one hop away from the end-host but > not quite into the network (and the Maelstrom work I presented day > before yesterday at NSDI would be one example). > > - mahesh > > -- > http://www.cs.cornell.edu/~mahesh > > > -----Original Message----- > From: end2end-interest-bounces at postel.org on behalf of David P. Reed > Sent: Fri 4/18/2008 9:13 AM > To: Jon Crowcroft > Cc: 'end2end-interest at postel.org' > Subject: Re: [e2e] end of interest > > I personally think that the network community has become frustrated with > the inability to explore end-to-end protocols because the endpoint > stacks are "locked in" by vendors in proprietary code. > > > From huitema at windows.microsoft.com Fri Apr 18 20:27:35 2008 From: huitema at windows.microsoft.com (Christian Huitema) Date: Fri, 18 Apr 2008 20:27:35 -0700 Subject: [e2e] end of interest In-Reply-To: References: <48089E95.1090103@reed.com> Message-ID: <8EFB68EAE061884A8517F2A755E8B60A02C7A22B05@NA-EXMSG-W601.wingroup.windeploy.ntdev.microsoft.com> In what world are you living, exactly? What about systems like Skype, or Bit torrent? They are definitely pushing the envelope of end to end designs, are widely deployed, and are not controlled by major corporations. From: end2end-interest-bounces at postel.org [mailto:end2end-interest-bounces at postel.org] On Behalf Of Mahesh Balakrishnan Sent: Friday, April 18, 2008 5:11 PM To: end2end-interest at postel.org Subject: Re: [e2e] end of interest In fact there seems to be pushback from both ends --- we can't deploy end-to-end protocols because major companies own the end-host stacks; and we can't push mechanisms deep into the network because ISPs and router companies own the network. Arguably the latter source of pushback played a major role in the emergence of the e2e philosophy; but now we have equally powerful commercial forces on the other side. So effectively the only practical mode of deployment seems to be the 'almost' end-to-end middlebox --- one hop away from the end-host but not quite into the network (and the Maelstrom work I presented day before yesterday at NSDI would be one example). - mahesh -- http://www.cs.cornell.edu/~mahesh -----Original Message----- From: end2end-interest-bounces at postel.org on behalf of David P. Reed Sent: Fri 4/18/2008 9:13 AM To: Jon Crowcroft Cc: 'end2end-interest at postel.org' Subject: Re: [e2e] end of interest I personally think that the network community has become frustrated with the inability to explore end-to-end protocols because the endpoint stacks are "locked in" by vendors in proprietary code. -------------- next part -------------- An HTML attachment was scrubbed... URL: http://mailman.postel.org/pipermail/end2end-interest/attachments/20080418/c3d17c27/attachment.html From ggm at apnic.net Fri Apr 18 20:57:38 2008 From: ggm at apnic.net (George Michaelson) Date: Sat, 19 Apr 2008 13:57:38 +1000 Subject: [e2e] end of interest In-Reply-To: References: Message-ID: <876CF2E2-0353-429F-851F-BD61D59B1ADF@apnic.net> I think the meme has forked. Into a three-legged trouser. One trouser-leg is about 'purer' network architectures where e2e is a feature, and requires no bumps in stacks, or cute tricks. In the moral spirit of the other great contribution from the late 20th C.. KISS, if you can have it, this is much nicer for the end/edge user. But the clear signal from the profit-centres-formally-known-as-telco is that you make less money in the apparent short term, by being nice. Many of us believe that both the social goodput, and the long-term profit would be better from structural-separation models without a need for bumps. This implies regulation. Despite many people's antithesis to regulation, some regulation appears useful, especially if you want to have structural separation and avoidance of e2e hindering bumps. The other leg is about the bumps and tricks (which probably would have made more sense as bumps and grinds..) you have to do, to work through NAT and other nasty intercepts on a 'pure' e2e network. The lesson from this side of the fork is that nothing is impossible, but the cost in packet and delay terms rises. At the crossing point where the cost exceeds some variable which depends on context, people either more to another supplier or give up. (ie 'live' video over tunneled SSH over HTTP over ICMP is of course always/sometimes possible, but gets un-rewarding. OTOH, if you were willing to wait for a torrent to download anyway. then almost nothing a network puts in your way will actually GET in your way, if you are sufficiently persistent.) The third trouserleg is where people observe that e2e birthed p2p, and find reasons to want to use it. At the point where Skype exploits your spare bandwidth as a virtual router cloud, torrents are used for distro's and now embraced by the mainstream as a viable distribution cloud and we're fighting in the FAA about 'network neutrality' at the same time, I think that e2e remains as interesting as ever. I think this list is about shaking the fluff out of the turn-ups, on the three legged trouser. A long time ago, on my fathers birthday, we got him a present 'for the man who has everything' which was a belly-button lint brush. perhaps thats also something useful for the e2e list? -George From alexsm at gmail.com Fri Apr 18 21:07:45 2008 From: alexsm at gmail.com (Alex Moura) Date: Sat, 19 Apr 2008 01:07:45 -0300 Subject: [e2e] SCTP popularity Message-ID: Greetings, Inspired by the last thread, I'd like to know your opinions regarding SCTP adoption and general use. It will never happen, or it will follow the same process - and time - that IPv6 adoption is taking? Best regards, Alex -------------- next part -------------- An HTML attachment was scrubbed... URL: http://mailman.postel.org/pipermail/end2end-interest/attachments/20080419/c430be18/attachment.html From ggm at apnic.net Fri Apr 18 22:04:05 2008 From: ggm at apnic.net (George Michaelson) Date: Sat, 19 Apr 2008 15:04:05 +1000 Subject: [e2e] SCTP popularity In-Reply-To: References: Message-ID: I thought it was interesting that ipfix mandated it as a transport. To the extent network droids care about measurement, and its going to be supported by vendors, and they are hooked on netflow and see ipfix as a logical successor, one can intuit that it will be enabled at the core unless they decide to disable it. FreeBSD trumpeted having 'the reference implementation' recently. There are times when that kind of claim is the kiss of death. Mandatory if-all-HQ-p0rn-was-on-SCTP-it-would-be-deployed-in-a-flash reference.. -George From mahesh at cs.cornell.edu Fri Apr 18 22:16:44 2008 From: mahesh at cs.cornell.edu (Mahesh Balakrishnan) Date: Sat, 19 Apr 2008 01:16:44 -0400 Subject: [e2e] end of interest References: <48089E95.1090103@reed.com> <8EFB68EAE061884A8517F2A755E8B60A02C7A22B05@NA-EXMSG-W601.wingroup.windeploy.ntdev.microsoft.com> Message-ID: In a world where clean-slate designs are impractical, apparently. I do belong to a generation of researchers who grew up on P2P and overlays --- and who accept the presence of IP and TCP as religious imperatives that must be engineered around. If industry realities confine us to particular philosophies of system design we can stop arguing about their relative merits and simply work with whatever is most practically deployable. Depending on your point of view this is either an excellent thing or the end of pure research as we know it. But I have to admit I don't understand how people can talk of designing the next-generation Internet from scratch when the economics of computer systems are so stacked against any kind of ground-up innovation. Also - if the router/ISP companies pushed us to the endpoints of the network and now the software companies have pushed us up the software stack --- where do we go next? - mahesh -----Original Message----- From: Christian Huitema [mailto:huitema at windows.microsoft.com] Sent: Fri 4/18/2008 11:27 PM To: Mahesh Balakrishnan; end2end-interest at postel.org Subject: RE: [e2e] end of interest In what world are you living, exactly? What about systems like Skype, or Bit torrent? They are definitely pushing the envelope of end to end designs, are widely deployed, and are not controlled by major corporations. From: end2end-interest-bounces at postel.org [mailto:end2end-interest-bounces at postel.org] On Behalf Of Mahesh Balakrishnan Sent: Friday, April 18, 2008 5:11 PM To: end2end-interest at postel.org Subject: Re: [e2e] end of interest In fact there seems to be pushback from both ends --- we can't deploy end-to-end protocols because major companies own the end-host stacks; and we can't push mechanisms deep into the network because ISPs and router companies own the network. Arguably the latter source of pushback played a major role in the emergence of the e2e philosophy; but now we have equally powerful commercial forces on the other side. So effectively the only practical mode of deployment seems to be the 'almost' end-to-end middlebox --- one hop away from the end-host but not quite into the network (and the Maelstrom work I presented day before yesterday at NSDI would be one example). - mahesh -- http://www.cs.cornell.edu/~mahesh -----Original Message----- From: end2end-interest-bounces at postel.org on behalf of David P. Reed Sent: Fri 4/18/2008 9:13 AM To: Jon Crowcroft Cc: 'end2end-interest at postel.org' Subject: Re: [e2e] end of interest I personally think that the network community has become frustrated with the inability to explore end-to-end protocols because the endpoint stacks are "locked in" by vendors in proprietary code. -------------- next part -------------- An HTML attachment was scrubbed... URL: http://mailman.postel.org/pipermail/end2end-interest/attachments/20080419/5ef2dc92/attachment.html From lars.eggert at nokia.com Sat Apr 19 08:05:16 2008 From: lars.eggert at nokia.com (Lars Eggert) Date: Sat, 19 Apr 2008 16:05:16 +0100 Subject: [e2e] end of interest In-Reply-To: <48095BA4.3020709@cs.utk.edu> References: <48089E95.1090103@reed.com> <48095BA4.3020709@cs.utk.edu> Message-ID: <2656536D-323E-405A-AD55-B6F6653FCDFF@nokia.com> On 2008-4-19, at 3:40, ext Keith Moore wrote: > Seems to me like the biggest barrier to deploying new L3/4 end-to- > end functionality isn't host stacks, but rather middleboxes that > don't know what to do with new IP packet types or protocol extensions. Exactly! Lars From lars.eggert at nokia.com Sat Apr 19 08:09:01 2008 From: lars.eggert at nokia.com (Lars Eggert) Date: Sat, 19 Apr 2008 16:09:01 +0100 Subject: [e2e] SCTP popularity In-Reply-To: References: Message-ID: On 2008-4-19, at 5:07, ext Alex Moura wrote: > Inspired by the last thread, I'd like to know your opinions > regarding SCTP > adoption and general use. It will never happen, or it will follow > the same > process - and time - that IPv6 adoption is taking? Adoption - sure; as George has pointed out, IPFIX is using it, and so do several other applications and application-layer protocols. General use as in by you and me behind our home NATs - not currently. But the Linux firewall/NAT has SCTP and DCCP support (and BSD has at least SCTP, I think), and if residential NAT vendors that use those systems allow enabling that functionality, it may happen. I grant you that that's a big if, because there is little incentive. Lars From Jon.Crowcroft at cl.cam.ac.uk Sat Apr 19 09:06:24 2008 From: Jon.Crowcroft at cl.cam.ac.uk (Jon Crowcroft) Date: Sat, 19 Apr 2008 17:06:24 +0100 Subject: [e2e] end of interest In-Reply-To: <2656536D-323E-405A-AD55-B6F6653FCDFF@nokia.com> References: <48089E95.1090103@reed.com> <48095BA4.3020709@cs.utk.edu> <2656536D-323E-405A-AD55-B6F6653FCDFF@nokia.com> Message-ID: so everyone talks about a bill of rights for spectrum and I've just seen a debate (partly triggered by the massive impact of the bbc's iplayer) on a bill of rights for p2p so maybe a more fundemental (e2e) idea is a bill of rights for packets what might this clarify? well, one of the middlebox key problems is that unlike typical IP receivers (e.g. routers or hosts), they don't obey postel's principle - this is (maybe) reasonable if you are being a firewall and trying to normalise things to some extend to reduce the load on an IDS or a host firewall but sometimes, these middlebox people get a bit TOO big for THEIR boots and impose rules that are arbitrary (viz the historical barrier to ecn deployment) what we need is a better security analysis of the middle classes to say what is reasonable to be suspicious, and what is just plain annoying.... of course it is far too late for this internet where we now see a class-ridden society (shades of python/cleese/barker/corbett sketch for brit-tv-philes: I look up to him because he has a permanent address. I look down on him because he is dhcpd and natted to death, but I look up to him because he has a /16 I look down on him because he is behind a firewall, whereas I have a /16 and am completely 0wned or something) actually, you know what would be neat would be a new network, where I make v6 sockaet API calls via RPCs over HTTP to a FQDN , so I can then do what I want, albeit slowly In missive <2656536D-323E-405A-AD55-B6F6653FCDFF at nokia.com>, Lars Eggert typed: >>On 2008-4-19, at 3:40, ext Keith Moore wrote: >>> Seems to me like the biggest barrier to deploying new L3/4 end-to- >>> end functionality isn't host stacks, but rather middleboxes that >>> don't know what to do with new IP packet types or protocol extensions. >> >>Exactly! >> >>Lars cheers jon From perfgeek at mac.com Sat Apr 19 09:51:50 2008 From: perfgeek at mac.com (rick jones) Date: Sat, 19 Apr 2008 09:51:50 -0700 Subject: [e2e] end of interest In-Reply-To: <480908CC.5050400@reed.com> References: <48089E95.1090103@reed.com> <480908CC.5050400@reed.com> Message-ID: On Apr 18, 2008, at 1:47 PM, David P. Reed wrote: > Andrew - > > I do exploration in Linux myself (I only run Windows in KVM, and > only when I absolutely have to). Regarding the Major Company's > proprietary TCP code - I agree that minor tweaks to the TCP standard > are adopted as long as the Major Company's programmers are the ones > who act as gatekeeprs. However, that's like what AT&T Bell Labs > used to say when they disparaged the Internet Experiment in the > 1980's - "when those lunatics working with Vint Cerf and Bob Kahn do > something really interesting, we'll put it in our network". Speaking as an employee, but not on behalf, of a Major Company with proprietary TCP code, I think that is exactly the way a Major Company does behave. And I don't see that as a Bad Thing (tm). While some of our customers might want the latest and greatest, by far most of them want stuff that Just Works and don't care to be experimental subjects. That means they don't want new stuff thrust upon them unbidden. The developers with Major Co may still put some new/ experimental things in the proprietary TCP code, but the bar will be quite high. "The network" (stack, intranet, whatnot) is much more critical/ valuable to Major Co's customers these days then say in '88/'89 when I and my colleagues were able to drop VJ Congestion control into MPE/XL TCP "just because it is better." It was also quite easy to demonstrate just how much better things were - and as part of that we (the developers at Major Co) were able to look at the research "those lunatics" :) published without using our customers as experimental subjects. The key today (and perhaps Catch-22) is to get the customers to request the features. > If you have no hope of deploying most innovations without bargaining > with Major Co., then why bother doing research? To demonstrate the the customers and employees of Major Co that the innovations are worthwhile. An irony in all this is that the customers of Major Co end-up demanding, and getting, changes in the TCP stacks that would probably leave some researchers aghast. Stuff like CKO, ACK avoidance heuristics, very configurable retransmission timers, etc... And miracle of miracle their networks are not known to have imploded :) rick jones there is no rest for the wicked, yet the virtuous have no pillows From lachlan.andrew at gmail.com Fri Apr 18 15:09:34 2008 From: lachlan.andrew at gmail.com (Lachlan Andrew) Date: Fri, 18 Apr 2008 15:09:34 -0700 Subject: [e2e] end of interest In-Reply-To: <480908CC.5050400@reed.com> References: <48089E95.1090103@reed.com> <480908CC.5050400@reed.com> Message-ID: Greetings David, On 18/04/2008, David P. Reed wrote: > I agree that minor tweaks to the TCP standard are adopted as long as the > Major Company's programmers are the ones who act as gatekeeprs. However, > that's like what AT&T Bell Labs used to say when they disparaged the > Internet Experiment in the 1980's > > All of this puts grit in the gears of innovation by requiring that > innovators "get permission" from those who want to preserve their market > power. Yes, but there is a big difference. AT&T wanted to regulate what other people did on AT&T's network, whereas the particular Major Company I was referring to is not telling other people what to do. It is just refraining from itself deploying technology that it has found to give its customers grief (like ECN or Window Scaling packets being dropped). > If you have no hope of deploying most innovations without bargaining with > Major Co., then why bother doing research? If it is truely end-to-end, then I don't think any Major Co. needs to be involved. Just ask the P2P community. Cheers, Lachlan -- Lachlan Andrew Dept of Computer Science, Caltech 1200 E California Blvd, Mail Code 256-80, Pasadena CA 91125, USA Ph: +1 (626) 395-8820 Fax: +1 (626) 568-3603 http://netlab.caltech.edu/lachlan From wyystars at gmail.com Fri Apr 18 04:41:40 2008 From: wyystars at gmail.com (wangyy) Date: Fri, 18 Apr 2008 19:41:40 +0800 Subject: [e2e] end of interest References: Message-ID: <000401c8a149$3090a710$d4cb6fa6@WANGYY> maybe end system will be added data-oriented layer, how to disperse data, get data, transport data, to balance or distract the load of end system. ----- Original Message ----- From: "Jon Crowcroft" To: ; Sent: Friday, April 18, 2008 6:10 PM Subject: [e2e] end of interest > last month's postings on this list was 80% CfPs; > this month has 1 message which is a CfP; > is this the end of interest in end-to-end? > > has the location/identifier debate > mean everyone split? > > does virtualization mean unjustifying the ends? > > has mesh networking for nanotech > or packet swarms and multipath routing > diverted attention elsewhere? > > its not like there's no work out there - just look at > this weeks NSDI programme:- > http://www.usenix.org/event/nsdi08/tech/ > > j. From dpreed at reed.com Sat Apr 19 12:14:11 2008 From: dpreed at reed.com (David P. Reed) Date: Sat, 19 Apr 2008 15:14:11 -0400 Subject: [e2e] end of interest In-Reply-To: <1208556801.23430.2.camel@sioux.systems.cs.cornell.edu> References: <48089E95.1090103@reed.com> <480908CC.5050400@reed.com> <1208556801.23430.2.camel@sioux.systems.cs.cornell.edu> Message-ID: <480A4483.8070703@reed.com> In the US and Europe at least, one Major Company that controls a network stack has been judged thoroughly and beyond appeal by the courts to have a legal monopoly, with the strong assertion that makes by definition about consequent market power. That *legal* position cannot be disputed. It would take a stronger argument than a mere vague handwave by a computer scientist toward the word "competing interests" to convince most economists and lawyers that when such a company keeps its network drivers protected, proprietary, and engages in agreements with hardware vendors to "certify" their drivers and hardware, the playing field for competition enables easy implementation of anything in that dominant network stack. Of course, computer scientists are welcome to their political opinions and dissent. But in science, dissent requires testable proof. Thus, I propose that the next PlanetLab scale experiment on new system architectures be carried out, not with Linux, but with Windows Vista. And without any prior agreement with Microsoft that gives the researchers licenses and access to code and internal interface privileges that students in, say, Ecuador don't have. Based on that test, we can ascertain whether the monopoly in legal fact has an impact on research freedom. Saikat Guha wrote: > On Fri, 2008-04-18 at 16:47 -0400, David P. Reed wrote: > >> If you have no hope of deploying most innovations without bargaining >> with Major Co., then why bother doing research? >> > > There are multiple major companies with competing interests ?though. > > Research that favors one would get deployed, or at least, would have a > chance of getting deployed. Google and network neutrality for example. > > From dpreed at reed.com Sat Apr 19 12:48:15 2008 From: dpreed at reed.com (David P. Reed) Date: Sat, 19 Apr 2008 15:48:15 -0400 Subject: [e2e] end of interest In-Reply-To: References: <48089E95.1090103@reed.com> <480908CC.5050400@reed.com> Message-ID: <480A4C7F.6080307@reed.com> Andrew - hello Lachlan Andrew wrote: > > AT&T wanted to regulate what > other people did on AT&T's network, whereas the particular Major > Company I was referring to is not telling other people what to do. It > is just refraining from itself deploying technology that it has found > to give its customers grief (like ECN or Window Scaling packets being > dropped). > > I have no evidence that Major Company always refrains from seeking competitive advantage by using its power to decide what gets bundled with its OS and what is certified to go with its OS. I note that AT&T always *claimed* it was "protecting the network from harm". And the same argument is often used by the technical experts of Major Company in explaining why, for example, they have to "break" compatibility with software products of strong applications competitors, or why, for example, hardware vendors must NOT include bundled software alternatives to that provided by Major Company when offering their products in the marketplace. It may even be true that there is a serious *risk* of harm. However, the Major Company is not God, and is not required to protect its customers from choice and the consequent risks. From day at std.com Sat Apr 19 16:12:13 2008 From: day at std.com (John Day) Date: Sat, 19 Apr 2008 19:12:13 -0400 Subject: [e2e] end of interest In-Reply-To: <480A4483.8070703@reed.com> References: <48089E95.1090103@reed.com> <480908CC.5050400@reed.com> <1208556801.23430.2.camel@sioux.systems.cs.cornell.edu> <480A4483.8070703@reed.com> Message-ID: Earlier Dave Reed said: "If you have no hope of deploying most innovations without bargaining with Major Co., then why bother doing research?" Ya know, I never knew that this was the primary purpose of research. How enlightening! Then more recently, Dave said: > >Based on that test, we can ascertain whether the monopoly in legal >fact has an impact on research freedom. Are there some rules I don't know about? They weren't taught in my school. I might remind our readers that in 1970 none of us were worried about whether or not AT&T would use what we were doing. Admittedly, we were being paid for our trouble, but most of us were just interested in what would happen and what we would learn from it. There is something amiss here. If the only reason for doing research is to see some large incumbent adopt it, then you either want another job or you aren't being sufficiently devious. ;-) Take care, John From touch at ISI.EDU Sat Apr 19 20:40:08 2008 From: touch at ISI.EDU (Joe Touch) Date: Sat, 19 Apr 2008 20:40:08 -0700 Subject: [e2e] end of interest In-Reply-To: References: <48089E95.1090103@reed.com> <48095BA4.3020709@cs.utk.edu> <2656536D-323E-405A-AD55-B6F6653FCDFF@nokia.com> Message-ID: <480ABB18.1060305@isi.edu> Jon Crowcroft wrote: > so everyone talks about a bill of rights for spectrum > and I've just seen a debate (partly triggered by > the massive impact of the bbc's iplayer) > on a bill of rights for p2p > > so maybe a more fundemental (e2e) idea is a bill of rights for packets See: http://www.isi.edu/touch/internet-rights/ Joe -------------- next part -------------- A non-text attachment was scrubbed... Name: signature.asc Type: application/pgp-signature Size: 250 bytes Desc: OpenPGP digital signature Url : http://mailman.postel.org/pipermail/end2end-interest/attachments/20080419/81d0ded8/signature.bin From touch at ISI.EDU Sat Apr 19 20:46:03 2008 From: touch at ISI.EDU (Joe Touch) Date: Sat, 19 Apr 2008 20:46:03 -0700 Subject: [e2e] end of interest In-Reply-To: References: Message-ID: <480ABC7B.8040208@isi.edu> Jon Crowcroft wrote: > last month's postings on this list was 80% CfPs; > this month has 1 message which is a CfP; > is this the end of interest in end-to-end? Just a note from your list admin: True, March was 80% CFPs. Feb was only 20%, and Jan was 3 out of 225 (1.33%). Overall, it just seems more like people were handling midterms, off on spring break, or otherwise temporarily occupied. I'll wait for that trend to persist for more than a month before I call for the paramedics, though discussion triggers are always welcome... Joe -------------- next part -------------- A non-text attachment was scrubbed... Name: signature.asc Type: application/pgp-signature Size: 250 bytes Desc: OpenPGP digital signature Url : http://mailman.postel.org/pipermail/end2end-interest/attachments/20080419/7c3bd169/signature.bin From day at std.com Sat Apr 19 21:17:19 2008 From: day at std.com (John Day) Date: Sun, 20 Apr 2008 00:17:19 -0400 Subject: [e2e] end of interest In-Reply-To: <12D20F05-721A-47B7-890E-C9DA143D9919@cisco.com> References: <48089E95.1090103@reed.com> <480908CC.5050400@reed.com> <1208556801.23430.2.camel@sioux.systems.cs.cornell.edu> <480A4483.8070703@reed.com> <12D20F05-721A-47B7-890E-C9DA143D9919@cisco.com> Message-ID: In the 1980s, people would ask me why I didn't know much about SNA. (For our younger readers, the then dominant data networking architecture.) I usually replied, that I had enough trouble remembering all the right ways to build networks, without trying to remember all the wrong ways too. That, in some sense, has always been what I have pursued. Never assume, always question, look for fundamentals, the invariants and be damned where it led. I wanted to know the "right" way regardless of whether IBM, Cisco, Microsoft, ITU, ISO, IEEE, or the IETF chose to adopt it. At the very least, I would have a better idea what kinds of compromises I was making when I had to deal with "reality." And who knows, you might find better ways of doing things. I guess this puts me in the rationalist camp. I have never been big on faith-based engineering (or science for that matter). Some may wish to emulate Clavius, more concerned with protecting the faith than going where the problem leads, that is not for me. Today, I look at the IETF projects; the vast majority of "research;" and my reaction has a lot in common with my reaction to SNA long ago. We take a 30 year old demo, patch and patch it, relying on Moore's Law to save us from the hard problems. a It bears a strong similarity to the late days of DOS, when what we need is Multics, but we would probably settle for Unix. Then we pat ourselves on the back for what brilliant engineers we are. When, in fact, we haven't done much at all. Sure there was some solid engineering and with Moore's Law we have built something that the world has found amazing. But as scientists, we have to be honest with ourselves. What have we done? What major new insights, breakthroughs, surprising results have we uncovered over the last 30 years? What have we done to consolidate, make it simpler? Get at the fundamentals. Ensure this thing is on a solid footing now that the world relies on it. We know what is really underneath. We could have used the public infatuation to buy ourselves some time, but it seems we have used it mostly to sun ourselves. Having successfully ignored it, we now have a 35 year old problem finally coming home to roost. So here we are when we need breakthroughs and all we can do is come up with lame half answers and discuss how many engineers it takes to clean a slate. And from the looks of it, we have forgotten how to do that too. Is this the problem of the Major Companies? Hardly. They are *suppose* to take a short term view and build product and keep customers satisfied. The problem has been with academic research that seems to have gotten the idea that university research is low risk angel funding, rather than pursuing a deeper understanding. Laying the groundwork for the future, so that these really hard problems don't bite the engineers in the future AND more importantly, that the engineers know the fundamentals so that when these hard problems finally come down to the engineering they know what the right answer is suppose to be. Guys, it is time that some of got back to doing real research, not just hot topics. Well this wasn't suppose to be a screed, but it seems to have turned into one! Ooops. ;-) Take care all, John Day At 19:15 -0700 2008/04/19, Fred Baker wrote: >On Apr 19, 2008, at 4:12 PM, John Day wrote: > >>I might remind our readers that in 1970 none of us were worried >>about whether or not AT&T would use what we were doing. >>Admittedly, we were being paid for our trouble, but most of us were >>just interested in what would happen and what we would learn from >>it. There is something amiss here. > >Thanks for saying that. I have been very disturbed by this thread. > >Cisco, my employer, and specifically the department I work in, funds >a fair bit of research into making things better. Reason: at the end >of the day, if things work better, people will buy more of our >products. Call that whatever you like; it is the fundamental >motivation of any Major Company to fund research. Now, if we're not >going to make things better, why fund the research? > >If your research isn't funded by a major company, then presumably >you hope that the outcome will be something different that Major >Company might have chosen. I came up the corporate side, not the >research side, but at the end of the day the success of whatever >company I worked for at the time depended on the notion that whoever >was the "Major Company" at the time was wrong. So far, I have been >mostly right. > >If your research was funded by a Major Company, then you are betting >that improvements to the present model are a good thing. >Paradoxically perhaps, thus far the things we have learned on this >axis have been useful as well. > >I'm seriously wondering where the bitter folks who hate the >businesses that made the Internet a worldwide communication >infrastructure are coming from. Do they hate the Internet? Would >they rather have an Internet that was undeployed for lack of >funding, and therefore the system they the say they dislike because >it is the last viable business standing? From Jon.Crowcroft at cl.cam.ac.uk Sun Apr 20 03:17:52 2008 From: Jon.Crowcroft at cl.cam.ac.uk (Jon Crowcroft) Date: Sun, 20 Apr 2008 11:17:52 +0100 Subject: [e2e] end of interest In-Reply-To: <6d2996bb0804200043i76986be9n32282fff2814ae02@mail.gmail.com> References: <48089E95.1090103@reed.com> <48095BA4.3020709@cs.utk.edu> <2656536D-323E-405A-AD55-B6F6653FCDFF@nokia.com> <480ABB18.1060305@isi.edu> <6d2996bb0804200043i76986be9n32282fff2814ae02@mail.gmail.com> Message-ID: cleanslate - i dont see that the virtualisation clean slate architectures violate any of Joe's internet rights at all - there's at least 3 virtualisation projects I can think of (washington st louis, princeton/gatech, and ucl/lancaster)... p2p v. isp: this is standard value chain stuff. the value of the net to users is that it connects them to content. the network providers are in the business of taking a fraction of the business that the content providers are in but it isn't good business to dictate to yor customer how they use what they pay for so the fact that iplayer does p2p delivery (and skype does p2p voip) is none of the ISPs business -in fact the cos to the consumer of the content service (bbc content, voip, etc) is massively redced by doing it this way, which means the profit to the content service provider is a percentage of a smalelr sum, and the percentage to be made by the common carrier (oops, sorry, ISP, so logn as it isnt comcast:) is smaller....forcing a content provider back to a data center (and more expensive (and less green)) alternative, increases their costs, which means the cost to the consumer goes up and the number of consumers goes down, and the number of boradband ISP customers goes down. so what do you want as an ISP - more higher bandwidth consumers with a lower margin, or less with a higher margin? do the math. (you need access to some guesstimates of the costs but for the bbc, its clear that the income is (in the uk) from license fee (plus sales of programs to the US)...if you ask them to run a massive data center (or pay akamai for distribution) then you will get zero customers. so the income from uk brodband custmers from that aspect goes to zero. so theres more incomei in providing it than not. ergo, the ISPs are posturing. In missive <6d2996bb0804200043i76986be9n32282fff2814ae02 at mail.gmail.com>, "Saikat Guha" typed: >>On Sat, Apr 19, 2008 at 11:40 PM, Joe Touch wrote: >>> > so maybe a more fundemental (e2e) idea is a bill of rights for packets >>> > >>> >>> See: >>> >>> http://www.isi.edu/touch/internet-rights/ >> >>Interesting; virtually all "clean-slate" architectures would violate >>the letter of these requirements, but perhaps not the spirit. >> >>More to the point of this discussion, point #4 reads "Users have the >>right to send [...] provided it uses an inconsequential amount of >>resources of the network". The debate between BBC's iPlayer and the UK >>ISPs seems to stem from the surge of traffic created (unilaterally) by >>the BBC's system designers. If the BBC coordinated with the ISPs, >>perhaps to place caches at appropriate locations, perhaps use P4P, >>etc., the ISP would have been better prepared to handle the surge. >> >>Does e2e tacitly encourage unilateralism in terms of ends vs. the >>middle, or is that an artifact of competing interests? >> >>-- >>Saikat cheers jon From Jon.Crowcroft at cl.cam.ac.uk Sun Apr 20 04:48:39 2008 From: Jon.Crowcroft at cl.cam.ac.uk (Jon Crowcroft) Date: Sun, 20 Apr 2008 12:48:39 +0100 Subject: [e2e] end of interest In-Reply-To: <6d2996bb0804200353p48c70166yf044678a7ab375f5@mail.gmail.com> References: <48089E95.1090103@reed.com> <48095BA4.3020709@cs.utk.edu> <2656536D-323E-405A-AD55-B6F6653FCDFF@nokia.com> <480ABB18.1060305@isi.edu> <6d2996bb0804200043i76986be9n32282fff2814ae02@mail.gmail.com> <6d2996bb0804200353p48c70166yf044678a7ab375f5@mail.gmail.com> Message-ID: In missive <6d2996bb0804200353p48c70166yf044678a7ab375f5 at mail.gmail.com>, "Saikat Guha" typed: >>On Sun, Apr 20, 2008 at 6:17 AM, Jon Crowcroft >> wrote: >>> cleanslate - i dont see that the virtualisation clean slate architectures violate any of Joe's >>> internet rights at all - there's at least 3 virtualisation projects I can think of >>> (washington st louis, princeton/gatech, and ucl/lancaster)... >>Fair enough. I was thinking more Triad/i3/DONA/defaultOff/... where >>strictly speaking, "DNS" and "IP" play a very different role than they >>do today. To the extent that the bill of Internet rights says: _users_ >>should be able to reach any other user at reasonable speeds (at some >>layer), I agree, but that's very different from saying all public IP >>all the time. totally agree - DONA (and schemes that start from no explicit sender or no explicit recipient - i.e. pub/sub systems which predate dona by 15 years which seem to be overlooked often in the data driven net arch post geni work:) also >>> iplayer does p2p delivery (and skype does p2p voip) is none of the ISPs >>> business [...] forcing a content provider back to a data center (and more >>> expensive (and less green)) alternative, increases their costs >>Skype is a good example of p2p creating unnecessary traffic in the >>sense that I shouldn't have to relay through Brazil for a peer 10kms >>away with a common ISP. A p4p-like approach where all parties have a >>control knob allow these inefficiencies to be removed, and perhaps >>forestalls unnecessary posturing. But it does require the content >>provider to convince the ISP to get on board. or else to tune their overlay of supereers by mapping the underlying topology - actualyl it would be unlikely to relay interactive voice traffic too far away coz of latency so it might be between bandwidth and latency estimation, overlay p2p superpeer node selection systems might do ok, modulo policies...:- what i'd like to see is an overlay ystem have an open API for programming in application layer polociy routes - a decent system (e.g. based on metarouting and xorp) might usefully be unified with the underlying BGP policies directly and so provide 21st century interdomain traffic engineering:) cheers jon From randy at psg.com Sun Apr 20 05:07:49 2008 From: randy at psg.com (Randy Bush) Date: Sun, 20 Apr 2008 21:07:49 +0900 Subject: [e2e] end of interest In-Reply-To: References: <48089E95.1090103@reed.com> <48095BA4.3020709@cs.utk.edu> <2656536D-323E-405A-AD55-B6F6653FCDFF@nokia.com> <480ABB18.1060305@isi.edu> <6d2996bb0804200043i76986be9n32282fff2814ae02@mail.gmail.com> Message-ID: <480B3215.3040008@psg.com> > p2p v. isp: this is standard value chain stuff. [ apologies, but you have triggered a rant i have been foaming about for some years ] this characterization, while historically true, is the root of a major mistake. big content owners (note that, unlike the internet model, they were not the content creators) helped create this polarization and lobby heavily to maintain and expand it as much as possible. in fact, it is bleedin' insane for isps to position against what the customer wants. if you can not provide the customer what they want and still keep some margin, then your business model is broken, you should not be in business, and soon will not be, unless you manage to get the lawyers and the government to tilt the billiards table. the isps, instead of making the p2p software authors think they have to work around and evade the topology, should help the p2p software discover how to most efficiently utilize the actual topology, to find the closest and most efficient neighbors, use the less congested paths, etc. this would be a win for the customer, who gets faster more efficient service, for the isp, whose bandwidth is used more efficiently, for the p2p software author, as their product looks darned good, and for the big content owners, who will be forced to enter the current century even sooner. some isps have heard this over the last few years, and are working with p2p authors. but i think this is mostly in asia, not the states or europe, or south. sorry to rant on. you pushed an old hot button. delete this message and spend the time reading john day's latest again. randy From jnc at mercury.lcs.mit.edu Sun Apr 20 07:56:37 2008 From: jnc at mercury.lcs.mit.edu (Noel Chiappa) Date: Sun, 20 Apr 2008 10:56:37 -0400 (EDT) Subject: [e2e] end of interest Message-ID: <20080420145637.8812E86AE4@mercury.lcs.mit.edu> > From: John Day > Well this wasn't suppose to be a screed, but it seems to have turned > into one! Ooops. Nah, good stuff. However, as someone who is fairly dedicated to the concept of "look for fundamentals, the invariants and be damned where it led", and who has also tried to get some of that thinking into the network, let me offer you a couple of decades worth of perspective from that particular seat. It's damned hard. There are in fact pretty good reasons for this. The old line about "updating the Wright flyer into a 747 - while it's flying" captures a lot of it. An unfinished note of mine (is there any other kind? :-) captures a slightly different take on it, which is that communication networks have a property which is not shared by editors, operating systems, etc, etc. That is that if you come up with the world's greatest new editor, if you get a few people to adopt it, they will love it, and more will join in, so eventually you get to, say, 2% market share, and that works just fine. Invent a new communication system, and get 2% of the market to use it, and ... very shortly you will have 0%. The reason is obvious: the point of communication systems is to communicate, and if you limit yourself to a small group of potential communicatees, your communication system doesn't have lot of perceived/actual value. A plain telephone that can reach everyone is a heck of a lot more valuable than a videophone that can reach 17 people. Which dumps us back into the Wright-flyer -> 747 problem ... which is not an easy one. Electric outlet sockets, light-bulb sockets, etc look much the same as they did a century ago - because interoperability with installed base is giant millstone. However, all is not lost. I think for a long time there have been understandable and rational reasons people were content to build more fabric wing. To start with, as the network grew, people were so ass-deep in alligators they couldn't take the time to do anything but build more fabric wing as fast as the sewing machines could crank it out. Then we went through a deep cutback where there wasn't any money to do anything anyway. However, I think that as the field has matured, and people have enough personal history in it to internalize some of these points, there is an increasing understanding out there of some of the things you're talking about, like: > We take a 30 year old demo, patch and patch it, relying on Moore's Law > to save us from the hard problems. > ... > Sure there was some solid engineering and with Moore's Law we have > built something that the world has found amazing. > ... > What have we done to consolidate, make it simpler? Get at the > fundamentals. Ensure this thing is on a solid footing now that the > world relies on it. And, moreover, I think (or perhaps this is just a hope that my desire is turning into a perception) that there's a growing appreciation of the *long-term value* of spending a little more resources now, to build something (and I'm talking about architecture here, not hardware) that has a little more flexibility and capability - and hence *durability*. Of course, we still have to make the numbers add up (both $ and engineering), but I think there's more interest now in 'doing the right thing'. Enough more to make it happen? Well, we'll see. But for those who are interested, I have a nice shiny-new metal wing concept for you.... > The problem has been with academic research that seems to have gotten > the idea that university research is low risk angel funding .. Guys, it > is time that some of got back to doing real research, not just hot > topics. Indeed. Some of the research I read about sure seems a lot like "how to build a better fabric wing with the latest hot-glue technology". Noel From fred at cisco.com Sat Apr 19 19:15:10 2008 From: fred at cisco.com (Fred Baker) Date: Sat, 19 Apr 2008 19:15:10 -0700 Subject: [e2e] end of interest In-Reply-To: References: <48089E95.1090103@reed.com> <480908CC.5050400@reed.com> <1208556801.23430.2.camel@sioux.systems.cs.cornell.edu> <480A4483.8070703@reed.com> Message-ID: <12D20F05-721A-47B7-890E-C9DA143D9919@cisco.com> On Apr 19, 2008, at 4:12 PM, John Day wrote: > I might remind our readers that in 1970 none of us were worried > about whether or not AT&T would use what we were doing. > Admittedly, we were being paid for our trouble, but most of us were > just interested in what would happen and what we would learn from > it. There is something amiss here. Thanks for saying that. I have been very disturbed by this thread. Cisco, my employer, and specifically the department I work in, funds a fair bit of research into making things better. Reason: at the end of the day, if things work better, people will buy more of our products. Call that whatever you like; it is the fundamental motivation of any Major Company to fund research. Now, if we're not going to make things better, why fund the research? If your research isn't funded by a major company, then presumably you hope that the outcome will be something different that Major Company might have chosen. I came up the corporate side, not the research side, but at the end of the day the success of whatever company I worked for at the time depended on the notion that whoever was the "Major Company" at the time was wrong. So far, I have been mostly right. If your research was funded by a Major Company, then you are betting that improvements to the present model are a good thing. Paradoxically perhaps, thus far the things we have learned on this axis have been useful as well. I'm seriously wondering where the bitter folks who hate the businesses that made the Internet a worldwide communication infrastructure are coming from. Do they hate the Internet? Would they rather have an Internet that was undeployed for lack of funding, and therefore the system they the say they dislike because it is the last viable business standing? From saikat at cs.cornell.edu Sun Apr 20 00:43:35 2008 From: saikat at cs.cornell.edu (Saikat Guha) Date: Sun, 20 Apr 2008 03:43:35 -0400 Subject: [e2e] end of interest In-Reply-To: <480ABB18.1060305@isi.edu> References: <48089E95.1090103@reed.com> <48095BA4.3020709@cs.utk.edu> <2656536D-323E-405A-AD55-B6F6653FCDFF@nokia.com> <480ABB18.1060305@isi.edu> Message-ID: <6d2996bb0804200043i76986be9n32282fff2814ae02@mail.gmail.com> On Sat, Apr 19, 2008 at 11:40 PM, Joe Touch wrote: > > so maybe a more fundemental (e2e) idea is a bill of rights for packets > > > > See: > > http://www.isi.edu/touch/internet-rights/ Interesting; virtually all "clean-slate" architectures would violate the letter of these requirements, but perhaps not the spirit. More to the point of this discussion, point #4 reads "Users have the right to send [...] provided it uses an inconsequential amount of resources of the network". The debate between BBC's iPlayer and the UK ISPs seems to stem from the surge of traffic created (unilaterally) by the BBC's system designers. If the BBC coordinated with the ISPs, perhaps to place caches at appropriate locations, perhaps use P4P, etc., the ISP would have been better prepared to handle the surge. Does e2e tacitly encourage unilateralism in terms of ends vs. the middle, or is that an artifact of competing interests? -- Saikat From saikat at cs.cornell.edu Sun Apr 20 03:53:28 2008 From: saikat at cs.cornell.edu (Saikat Guha) Date: Sun, 20 Apr 2008 06:53:28 -0400 Subject: [e2e] end of interest In-Reply-To: References: <48089E95.1090103@reed.com> <48095BA4.3020709@cs.utk.edu> <2656536D-323E-405A-AD55-B6F6653FCDFF@nokia.com> <480ABB18.1060305@isi.edu> <6d2996bb0804200043i76986be9n32282fff2814ae02@mail.gmail.com> Message-ID: <6d2996bb0804200353p48c70166yf044678a7ab375f5@mail.gmail.com> On Sun, Apr 20, 2008 at 6:17 AM, Jon Crowcroft wrote: > cleanslate - i dont see that the virtualisation clean slate architectures violate any of Joe's > internet rights at all - there's at least 3 virtualisation projects I can think of > (washington st louis, princeton/gatech, and ucl/lancaster)... Fair enough. I was thinking more Triad/i3/DONA/defaultOff/... where strictly speaking, "DNS" and "IP" play a very different role than they do today. To the extent that the bill of Internet rights says: _users_ should be able to reach any other user at reasonable speeds (at some layer), I agree, but that's very different from saying all public IP all the time. > iplayer does p2p delivery (and skype does p2p voip) is none of the ISPs > business [...] forcing a content provider back to a data center (and more > expensive (and less green)) alternative, increases their costs Skype is a good example of p2p creating unnecessary traffic in the sense that I shouldn't have to relay through Brazil for a peer 10kms away with a common ISP. A p4p-like approach where all parties have a control knob allow these inefficiencies to be removed, and perhaps forestalls unnecessary posturing. But it does require the content provider to convince the ISP to get on board. -- Saikat From touch at ISI.EDU Sun Apr 20 09:15:31 2008 From: touch at ISI.EDU (Joe Touch) Date: Sun, 20 Apr 2008 09:15:31 -0700 Subject: [e2e] end of interest In-Reply-To: References: <48089E95.1090103@reed.com> <48095BA4.3020709@cs.utk.edu> <2656536D-323E-405A-AD55-B6F6653FCDFF@nokia.com> <480ABB18.1060305@isi.edu> <6d2996bb0804200043i76986be9n32282fff2814ae02@mail.gmail.com> Message-ID: <480B6C23.8090709@isi.edu> Jon Crowcroft wrote: > cleanslate - i dont see that the virtualisation clean slate architectures violate any of Joe's > internet rights at all - there's at least 3 virtualisation projects I can think of > (washington st louis, princeton/gatech, and ucl/lancaster)... And mine (X-Bone) - which motivated it, which wasn't clean slate because it didn't need to be ;-) Joe -------------- next part -------------- A non-text attachment was scrubbed... Name: signature.asc Type: application/pgp-signature Size: 250 bytes Desc: OpenPGP digital signature Url : http://mailman.postel.org/pipermail/end2end-interest/attachments/20080420/4a102ea2/signature.bin From day at std.com Sun Apr 20 19:52:11 2008 From: day at std.com (John Day) Date: Sun, 20 Apr 2008 22:52:11 -0400 Subject: [e2e] end of interest In-Reply-To: <20080420145637.8812E86AE4@mercury.lcs.mit.edu> References: <20080420145637.8812E86AE4@mercury.lcs.mit.edu> Message-ID: At 10:56 -0400 2008/04/20, Noel Chiappa wrote: > > From: John Day > > > Well this wasn't suppose to be a screed, but it seems to have turned > > into one! Ooops. > >Nah, good stuff. Thank you. > >However, as someone who is fairly dedicated to the concept of "look for >fundamentals, the invariants and be damned where it led", and who has also >tried to get some of that thinking into the network, let me offer you a >couple of decades worth of perspective from that particular seat. I was making a different point. There is a difference between doing networking research and doing work engineering the Internet. Today there seems to be only engineering the Internet, not much if any research on networking. There are many kinds internet-like networks possible. The Internet is just the one we kludged up some years back. We need to have some idea of the "answer" is regardless of whether the Internet ever reaches it. To do otherwise is, in effect, wandering lost in the woods. And as everyone knows, if you are lost in the woods, it is better to stay put and let rescuers come to you. It is very unlikely that you will FIND your way out. > >It's damned hard. > >There are in fact pretty good reasons for this. The old line about "updating >the Wright flyer into a 747 - while it's flying" captures a lot of it. One historical note of slight relevance: the methods of the Wright flyer do not serve as a basis for the 747, as much as the Curtiss flyer does. > >An unfinished note of mine (is there any other kind? :-) captures a slightly >different take on it, which is that communication networks have a property >which is not shared by editors, operating systems, etc, etc. That is that if >you come up with the world's greatest new editor, if you get a few people to >adopt it, they will love it, and more will join in, so eventually you get to, >say, 2% market share, and that works just fine. Invent a new communication >system, and get 2% of the market to use it, and ... very shortly you will >have 0%. The reason is obvious: the point of communication systems is to >communicate, and if you limit yourself to a small group of potential >communicatees, your communication system doesn't have lot of perceived/actual >value. A plain telephone that can reach everyone is a heck of a lot more >valuable than a videophone that can reach 17 people. > >Which dumps us back into the Wright-flyer -> 747 problem ... which is not an >easy one. Electric outlet sockets, light-bulb sockets, etc look much the same >as they did a century ago - because interoperability with installed base is >giant millstone. All of these examples represent an end. I for one do not believe we are anywhere near such an end. We haven't even begun to explore truly distributed systems, regardless of that the poor peer-to-peer [sic] advocates may think. The current architecture makes it very difficult to even contemplate such applications. There is much more that can be done, once we get around this particular millstone. > >However, all is not lost. > > >I think for a long time there have been understandable and rational reasons >people were content to build more fabric wing. To start with, as the network >grew, people were so ass-deep in alligators they couldn't take the time to do >anything but build more fabric wing as fast as the sewing machines could crank >it out. Then we went through a deep cutback where there wasn't any money to do >anything anyway. The engineers were, but they weren't the ones who should have been clearing brush. There were lots of people working on this and not everyone had to be ass-deep in alligators. We have known what the fundamental issues were for 30 years. But 2nd generation effect took over. The funny thing is that most of what needed to be done required little, if any, money. It did require some hard thinking, but of course spending money to build things is always more fun than hard thinking. > >However, I think that as the field has matured, and people have enough >personal history in it to internalize some of these points, there is an >increasing understanding out there of some of the things you're talking about, >like: > > > We take a 30 year old demo, patch and patch it, relying on Moore's Law > > to save us from the hard problems. > > ... > > Sure there was some solid engineering and with Moore's Law we have > > built something that the world has found amazing. > > ... > > What have we done to consolidate, make it simpler? Get at the > > fundamentals. Ensure this thing is on a solid footing now that the > > world relies on it. > >And, moreover, I think (or perhaps this is just a hope that my desire is >turning into a perception) that there's a growing appreciation of the >*long-term value* of spending a little more resources now, to build something >(and I'm talking about architecture here, not hardware) that has a little >more flexibility and capability - and hence *durability*. Frankly, until there is a group of people who see the Internet as we saw the phone company it won't happen. Too many people are too worried about protecting their legacy than furthering the understanding. To much calling for a revolution as long as you don't change what they did. (Some revolution.) Until they really see the fundamental flaws that are holding them back, it won't happen. (In fact it doesn't speak well for them that they haven't seen them already. They are so blatant!) > >Of course, we still have to make the numbers add up (both $ and engineering), >but I think there's more interest now in 'doing the right thing'. Enough more >to make it happen? Well, we'll see. But for those who are interested, I have >a nice shiny-new metal wing concept for you.... Which is? > > > > The problem has been with academic research that seems to have gotten > > the idea that university research is low risk angel funding .. Guys, it > > is time that some of got back to doing real research, not just hot > > topics. > >Indeed. Some of the research I read about sure seems a lot like "how to build >a better fabric wing with the latest hot-glue technology". Yea, I am afraid there needs to be lessons in how to clean a slate. ;-) They sure are coming out with a lot of dust on them. Take care, John -------------- next part -------------- An HTML attachment was scrubbed... URL: http://mailman.postel.org/pipermail/end2end-interest/attachments/20080420/2dec5fd2/attachment.html From rsofia at inescporto.pt Mon Apr 21 02:34:06 2008 From: rsofia at inescporto.pt (Rute Sofia) Date: Mon, 21 Apr 2008 10:34:06 +0100 Subject: [e2e] end of interest In-Reply-To: <8EFB68EAE061884A8517F2A755E8B60A02C7A22B05@NA-EXMSG-W601.wingroup.windeploy.ntdev.microsoft.com> References: <48089E95.1090103@reed.com> <8EFB68EAE061884A8517F2A755E8B60A02C7A22B05@NA-EXMSG-W601.wingroup.windeploy.ntdev.microsoft.com> Message-ID: <480C5F8E.6000903@inescporto.pt> Agree with Christian. And specifically focusing on L3, there seems to be other wave coming, with FON and now the more recent Whisher. New business roles (virtual ISPs, micro-ISPs) are also emerging. Rute Christian Huitema wrote: > In what world are you living, exactly? What about systems like Skype, or > Bit torrent? They are definitely pushing the envelope of end to end > designs, are widely deployed, and are not controlled by major corporations. > > > > *From:* end2end-interest-bounces at postel.org > [mailto:end2end-interest-bounces at postel.org] *On Behalf Of *Mahesh > Balakrishnan > *Sent:* Friday, April 18, 2008 5:11 PM > *To:* end2end-interest at postel.org > *Subject:* Re: [e2e] end of interest > > > > In fact there seems to be pushback from both ends --- we can't deploy > end-to-end protocols because major companies own the end-host stacks; > and we can't push mechanisms deep into the network because ISPs and > router companies own the network. Arguably the latter source of pushback > played a major role in the emergence of the e2e philosophy; but now we > have equally powerful commercial forces on the other side. > > So effectively the only practical mode of deployment seems to be the > 'almost' end-to-end middlebox --- one hop away from the end-host but not > quite into the network (and the Maelstrom work I presented day before > yesterday at NSDI would be one example). > > - mahesh > > -- > http://www.cs.cornell.edu/~mahesh > > > -----Original Message----- > From: end2end-interest-bounces at postel.org on behalf of David P. Reed > Sent: Fri 4/18/2008 9:13 AM > To: Jon Crowcroft > Cc: 'end2end-interest at postel.org' > Subject: Re: [e2e] end of interest > > I personally think that the network community has become frustrated with > the inability to explore end-to-end protocols because the endpoint > stacks are "locked in" by vendors in proprietary code. > > > -- Rute Sofia, PhD (rsofia at inescporto.pt, +351 22 2094263) http://www.inescporto.pt/utm/~ian/ Area Leader/Responsa'vel de A'rea IAN: Internet Architectures and Networking UTM: Telecommunications and Networking Unit INESC Porto From Jon.Crowcroft at cl.cam.ac.uk Mon Apr 21 04:21:39 2008 From: Jon.Crowcroft at cl.cam.ac.uk (Jon Crowcroft) Date: Mon, 21 Apr 2008 12:21:39 +0100 Subject: [e2e] end of interest In-Reply-To: <12D20F05-721A-47B7-890E-C9DA143D9919@cisco.com> References: <48089E95.1090103@reed.com> <480908CC.5050400@reed.com> <1208556801.23430.2.camel@sioux.systems.cs.cornell.edu> <480A4483.8070703@reed.com> <12D20F05-721A-47B7-890E-C9DA143D9919@cisco.com> Message-ID: I wouldn't be so worried for this reason but a more complex one I havnt seen any eveidence that having a world dominated by microsoft client systems, unix server systems and cisco-compliant router systems has caused problems for the uptake of any decent ideas in layes 3 or 4 the problem is that we don't have a 2 stakeholder system anymore - george michaelson hinted it when saying the meme has forked ('Into a three-legged trouser.') the problem is that we have end systems, intermediate layer 2/3 systems and intermediate layer 3-7 systems, and so there are THREE stakeholders in any change to the end-to-end model - anyone knows that three is a very stable (or ossified) configuration. actually, going back in time, I seem to dimly recall that once upon a time there was only 1 system (somethign that could send, receive or forward IP packets, e.g. according to joe touch's mantra)....then there were two, as the market split the world into sun and cisco, then microsoft woke up....etc etc but the key thing here is that we need to kill off one of the systems to regain any chance of anyone (even a big player) changing anything... I don't see how to kill off middleboxes as they protect hosts from bad guys (allegedly) so the only solution is to kill off routers and replace them entirely with middleboxes.... errr...oops, that can't be right :-) In missive <12D20F05-721A-47B7-890E-C9DA143D9919 at cisco.com>, Fred Bake r typed: >> >>On Apr 19, 2008, at 4:12 PM, John Day wrote: >> >>> I might remind our readers that in 1970 none of us were worried >>> about whether or not AT&T would use what we were doing. >>> Admittedly, we were being paid for our trouble, but most of us were >>> just interested in what would happen and what we would learn from >>> it. There is something amiss here. >> >>Thanks for saying that. I have been very disturbed by this thread. >> >>Cisco, my employer, and specifically the department I work in, funds >>a fair bit of research into making things better. Reason: at the end >>of the day, if things work better, people will buy more of our >>products. Call that whatever you like; it is the fundamental >>motivation of any Major Company to fund research. Now, if we're not >>going to make things better, why fund the research? >> >>If your research isn't funded by a major company, then presumably you >>hope that the outcome will be something different that Major Company >>might have chosen. I came up the corporate side, not the research >>side, but at the end of the day the success of whatever company I >>worked for at the time depended on the notion that whoever was the >>"Major Company" at the time was wrong. So far, I have been mostly right. >> >>If your research was funded by a Major Company, then you are betting >>that improvements to the present model are a good thing. >>Paradoxically perhaps, thus far the things we have learned on this >>axis have been useful as well. >> >>I'm seriously wondering where the bitter folks who hate the >>businesses that made the Internet a worldwide communication >>infrastructure are coming from. Do they hate the Internet? Would they >>rather have an Internet that was undeployed for lack of funding, and >>therefore the system they the say they dislike because it is the last >>viable business standing? cheers jon From dpreed at reed.com Mon Apr 21 06:52:55 2008 From: dpreed at reed.com (David P. Reed) Date: Mon, 21 Apr 2008 09:52:55 -0400 Subject: [e2e] end of interest In-Reply-To: References: <48089E95.1090103@reed.com> <480908CC.5050400@reed.com> Message-ID: <480C9C37.40509@reed.com> Rick - I'm clearly not explaining myself. rick jones wrote: > > "The network" (stack, intranet, whatnot) is much more > critical/valuable to Major Co's customers these days then say in > '88/'89 when I and my colleagues were able to drop VJ Congestion > control into MPE/XL TCP "just because it is better." It was also > quite easy to demonstrate just how much better things were - and as > part of that we (the developers at Major Co) were able to look at the > research "those lunatics" :) published without using our customers as > experimental subjects. > I was certainly not suggesting that any Major Company should experiment on its customers. If you read my point carefully, it was suggesting that Major Company has gone much farther than that (as did AT&T). It is prohibiting or massively limiting experimentation *by* its customers - which now includes the research community. And further, there are entities (perhaps not Major Company itself) who believe that experimentation at scale is highly dangerous. So dangerous, it must be banned, blocked, and treated as harmful, no matter what experiment is involved. This banning/blocking/raising alarms is part of a pattern that has a side effect of requiring that all innovators must buy permission from Major Company to do experimentation in a large area of potential interest. It also blocks new "start ups" that would compete with Major Company, but I will leave that one for the political arena here. I am focused only on the potential for free scientific experimentation. Now if the risk were truly of the form where society suggests we require level P4 confinement in biological labs (trenches with biopoisons around negative pressure chambers, only biohazard suits allowed), I think the crowd of "security zealots" who feel we must ban all variants of TCP from all clients, and definitely ban non-compliant IP protocol numbers might have a point. But such risks don't seem to be very likely. Leading me to be suspicious that a large part of the problem of Major Company's market power and unwillingness to allow free experimentation with the lower level of its network stacks is less focused on protecting *its* customers, and more focused on extending control over a larger and larger field of endeavor. In behaving this way, it is acting as many claim any profit-making company should - not morally, not based on social good, not even based on doing "good science". Just profit for its investors. This need not be the case for its employees - many are very good people. But they don't get rewarded *by the company* for being good people. They get rewarded *only* for making more money for the investors. And while sad, I don't personally believe that Google, Microsoft, or any company is capable of "not being evil", nor are they capable of "supporting good science". It's not their nature. If you work for them (and I do, since some of my work is sponsored by them), it's important to accept that any focus on social good is one's personal responsibility, and company requirements do not give one an "out" or make one's work safe from criticism for not recognizing the limits the company's requirements place on doing "good science" or "socially responsible engineering". Our professional organizations (IEEE and ACM) place requirements on us, *as professionals*, that go beyond our corporate masters' natural and amoral state of greed. The word "professional" is a class recognized by society as having a legal responsibility beyond drawing a paycheck and making money. From dpreed at reed.com Mon Apr 21 07:09:14 2008 From: dpreed at reed.com (David P. Reed) Date: Mon, 21 Apr 2008 10:09:14 -0400 Subject: [e2e] end of interest In-Reply-To: <12D20F05-721A-47B7-890E-C9DA143D9919@cisco.com> References: <48089E95.1090103@reed.com> <480908CC.5050400@reed.com> <1208556801.23430.2.camel@sioux.systems.cs.cornell.edu> <480A4483.8070703@reed.com> <12D20F05-721A-47B7-890E-C9DA143D9919@cisco.com> Message-ID: <480CA00A.2060204@reed.com> As a non-bitter person, I have to answer Fred below. Fred Baker wrote: > > If your research was funded by a Major Company, then you are betting > that improvements to the present model are a good thing. Paradoxically > perhaps, thus far the things we have learned on this axis have been > useful as well. Not true - major companies often fund alternatives to its present model - even internally. Perhaps Cisco doesn't fund investigations of what it is NOT selling? In fact, I personally think the only companies who can *afford* not to study alternatives are those who are comfortably able to enforce their market power against innovators. In fact, piling on some criticism of current practices in some companies, funding universities to do what the company thinks might be merely small, incremental improvements just turns universities into outsourced development shops for graduate students to do work for the companies that the companies can't afford to pay full time employees to do. Research is quite different from contract programming. > > I'm seriously wondering where the bitter folks who hate the businesses > that made the Internet a worldwide communication infrastructure are > coming from. Do they hate the Internet? Would they rather have an > Internet that was undeployed for lack of funding, and therefore the > system they the say they dislike because it is the last viable > business standing? Perhaps you misread. I don't hate businesses - I worked too long and proudly in industry R&D to hate them. I highly respect businesses and their power. I am hardly bitter as a result, and in fact I love the Internet. I don't think Cisco or BBN or Microsoft or Google are humans, and 99.9% of their current employees, if you think about it, had little to do with the creative acts that created the Internet. Those *organizations* are not worthy of thanks or worship. They are worthy of respect, but also demand criticism - criticism EQUAL to their power today. I do feel free to criticize the companies and their policies. In fact, I feel that the paragraph above is insulting, twisted, and just plain wrong: Ascribing motives that are not there (bitter, hate, ...) is rude, obnoxious, and yet highly typical of the way discourse plays out on FOX, CNN, MSNBC in politics. I'm tempted to utter expletives, Fred. But I won't. From Jon.Crowcroft at cl.cam.ac.uk Mon Apr 21 07:09:14 2008 From: Jon.Crowcroft at cl.cam.ac.uk (Jon Crowcroft) Date: Mon, 21 Apr 2008 15:09:14 +0100 Subject: [e2e] experimenting on customers In-Reply-To: <480C9C37.40509@reed.com> References: <48089E95.1090103@reed.com> <480908CC.5050400@reed.com> <480C9C37.40509@reed.com> Message-ID: I dont understand all this - most software in the last 30 years is an experiment on customers - the internet as a whole is an experiment the fact is that any sufficiently large and complex system that we cannot explain is necessarily not somethign where you can say anythign certain about what you are selling - ok so it isn't quite snake oil, but it sure aint like selling tables and chairs either so if we are honest, we'd admit this and say what we need is a pharma model of informed consent yeah, even discounts - i would happily run experimental (e.g. ipv6 or tcp-multihome) code for a reduction in the price of my internet access:) indeed, our cousins in the gaming industry PAY people to sit thru endless hours of fun on prototype games... In missive <480C9C37.40509 at reed.com>, "David P. Reed" typed: >>Rick - I'm clearly not explaining myself. >> >>rick jones wrote: >>> >>> "The network" (stack, intranet, whatnot) is much more >>> critical/valuable to Major Co's customers these days then say in >>> '88/'89 when I and my colleagues were able to drop VJ Congestion >>> control into MPE/XL TCP "just because it is better." It was also >>> quite easy to demonstrate just how much better things were - and as >>> part of that we (the developers at Major Co) were able to look at the >>> research "those lunatics" :) published without using our customers as >>> experimental subjects. >>> >>I was certainly not suggesting that any Major Company should experiment >>on its customers. If you read my point carefully, it was suggesting >>that Major Company has gone much farther than that (as did AT&T). It >>is prohibiting or massively limiting experimentation *by* its customers >>- which now includes the research community. >> >>And further, there are entities (perhaps not Major Company itself) who >>believe that experimentation at scale is highly dangerous. So >>dangerous, it must be banned, blocked, and treated as harmful, no matter >>what experiment is involved. >> >>This banning/blocking/raising alarms is part of a pattern that has a >>side effect of requiring that all innovators must buy permission from >>Major Company to do experimentation in a large area of potential interest. >> >>It also blocks new "start ups" that would compete with Major Company, >>but I will leave that one for the political arena here. I am focused >>only on the potential for free scientific experimentation. >> >>Now if the risk were truly of the form where society suggests we require >>level P4 confinement in biological labs (trenches with biopoisons around >>negative pressure chambers, only biohazard suits allowed), I think the >>crowd of "security zealots" who feel we must ban all variants of TCP >>from all clients, and definitely ban non-compliant IP protocol numbers >>might have a point. >> >>But such risks don't seem to be very likely. Leading me to be >>suspicious that a large part of the problem of Major Company's market >>power and unwillingness to allow free experimentation with the lower >>level of its network stacks is less focused on protecting *its* >>customers, and more focused on extending control over a larger and >>larger field of endeavor. >> >>In behaving this way, it is acting as many claim any profit-making >>company should - not morally, not based on social good, not even based >>on doing "good science". Just profit for its investors. This need not >>be the case for its employees - many are very good people. But they >>don't get rewarded *by the company* for being good people. They get >>rewarded *only* for making more money for the investors. >> >>And while sad, I don't personally believe that Google, Microsoft, or any >>company is capable of "not being evil", nor are they capable of >>"supporting good science". It's not their nature. If you work for them >>(and I do, since some of my work is sponsored by them), it's important >>to accept that any focus on social good is one's personal >>responsibility, and company requirements do not give one an "out" or >>make one's work safe from criticism for not recognizing the limits the >>company's requirements place on doing "good science" or "socially >>responsible engineering". >> >>Our professional organizations (IEEE and ACM) place requirements on us, >>*as professionals*, that go beyond our corporate masters' natural and >>amoral state of greed. The word "professional" is a class recognized by >>society as having a legal responsibility beyond drawing a paycheck and >>making money. cheers jon From dpreed at reed.com Mon Apr 21 07:16:53 2008 From: dpreed at reed.com (David P. Reed) Date: Mon, 21 Apr 2008 10:16:53 -0400 Subject: [e2e] end of interest In-Reply-To: <6d2996bb0804200043i76986be9n32282fff2814ae02@mail.gmail.com> References: <48089E95.1090103@reed.com> <48095BA4.3020709@cs.utk.edu> <2656536D-323E-405A-AD55-B6F6653FCDFF@nokia.com> <480ABB18.1060305@isi.edu> <6d2996bb0804200043i76986be9n32282fff2814ae02@mail.gmail.com> Message-ID: <480CA1D5.10209@reed.com> Saikat Guha wrote: > On Sat, Apr 19, 2008 at 11:40 PM, Joe Touch wrote: > >> >> http://www.isi.edu/touch/internet-rights/ >> > > Interesting; virtually all "clean-slate" architectures would violate > the letter of these requirements, but perhaps not the spirit. > > Perhaps not surprising at all - one point of "clean slate" as a research agenda was to consider alternatives to the Internet's design principles! I'm personally really uncomfortable about confusing research agendas with ideas such as "rights", "requirements", and "spirit". The latter terms seem to be about the Law, Product Engineering, and Law/Religion respectively. Research (I agree with John Day on this!) is about formulating questions and answering them. From dpreed at reed.com Mon Apr 21 07:26:03 2008 From: dpreed at reed.com (David P. Reed) Date: Mon, 21 Apr 2008 10:26:03 -0400 Subject: [e2e] end of interest In-Reply-To: <480B3215.3040008@psg.com> References: <48089E95.1090103@reed.com> <48095BA4.3020709@cs.utk.edu> <2656536D-323E-405A-AD55-B6F6653FCDFF@nokia.com> <480ABB18.1060305@isi.edu> <6d2996bb0804200043i76986be9n32282fff2814ae02@mail.gmail.com> <480B3215.3040008@psg.com> Message-ID: <480CA3FB.4040200@reed.com> Randy - this was a theme in both my and Dave Clark's testimony at the FCC en banc on reasonable network management practices at Harvard. Both of us suggested that if ISPs would open up the discussion of whatever problems P2P systems cause for their access networks, the solutions could be developed in an open transparent process (probably in front of the entire world at IETF, as other solutions have been created in the past). But instead of data about traffic in ISP's networks, we get slideware from Ellacoya and other vendors of amelioration systems that claim to show surges of "video" or "P2P", but when asked where the data comes from, all we get is vague allusions to "companies won't let us share" but "it's typical". As a scientist, I'm trained to be skeptical of claims without reproducibility. Not all my peers in the network field are scientists, I realize. Many are just selling stuff they happen to have to sell. P2P is a great bugbear - you can conjure sales with it. And you can recruit the kiddie porn and copyright protection lobbies to add to sales and political pressure. But scientifically, it's just an interesting problem that is worth solving - if you can reproduce and test your solutions. Randy Bush wrote: >> p2p v. isp: this is standard value chain stuff. >> > > [ apologies, but you have triggered a rant i have been foaming about for > some years ] > > this characterization, while historically true, is the root of a major > mistake. big content owners (note that, unlike the internet model, they > were not the content creators) helped create this polarization and lobby > heavily to maintain and expand it as much as possible. > > in fact, it is bleedin' insane for isps to position against what the > customer wants. if you can not provide the customer what they want and > still keep some margin, then your business model is broken, you should > not be in business, and soon will not be, unless you manage to get the > lawyers and the government to tilt the billiards table. > > the isps, instead of making the p2p software authors think they have to > work around and evade the topology, should help the p2p software > discover how to most efficiently utilize the actual topology, to find > the closest and most efficient neighbors, use the less congested paths, etc. > > this would be a win for the customer, who gets faster more efficient > service, for the isp, whose bandwidth is used more efficiently, for the > p2p software author, as their product looks darned good, and for the big > content owners, who will be forced to enter the current century even sooner. > > some isps have heard this over the last few years, and are working with > p2p authors. but i think this is mostly in asia, not the states or > europe, or south. > > sorry to rant on. you pushed an old hot button. delete this message > and spend the time reading john day's latest again. > > randy > > From randy at psg.com Mon Apr 21 07:33:45 2008 From: randy at psg.com (Randy Bush) Date: Mon, 21 Apr 2008 23:33:45 +0900 Subject: [e2e] end of interest In-Reply-To: <480CA3FB.4040200@reed.com> References: <48089E95.1090103@reed.com> <48095BA4.3020709@cs.utk.edu> <2656536D-323E-405A-AD55-B6F6653FCDFF@nokia.com> <480ABB18.1060305@isi.edu> <6d2996bb0804200043i76986be9n32282fff2814ae02@mail.gmail.com> <480B3215.3040008@psg.com> <480CA3FB.4040200@reed.com> Message-ID: <480CA5C9.6090703@psg.com> > But scientifically, it's just an interesting problem that is worth > solving - if you can reproduce and test your solutions. p2p is not a problem. it is a primitive solution to some social problems and some technical ones. some of the p2p protocols are even kinda distributed, woo woo! the first engineering problem is how to leak topo and congestion info to p2p services without exposing my network to attacks. what kind of info can i safely leak? that'll give us a clue as to what info the p2pware might have so it can be smarter about distribution. randy From mahesh at cs.cornell.edu Mon Apr 21 07:34:57 2008 From: mahesh at cs.cornell.edu (Mahesh Balakrishnan) Date: Mon, 21 Apr 2008 10:34:57 -0400 Subject: [e2e] end of interest References: <48089E95.1090103@reed.com> <8EFB68EAE061884A8517F2A755E8B60A02C7A22B05@NA-EXMSG-W601.wingroup.windeploy.ntdev.microsoft.com> <480C5F8E.6000903@inescporto.pt> Message-ID: Is there confusion, perhaps, between e2e as a philosophy and e2e as a necessity? I agree with Christian that a lot of the exciting new stuff happening is end-to-end. However: bit-torrent is e2e not because there are clean fundamental reasons for it to be so (though those may exist), but because that's the only deployment model open to the designer. I am not arguing that it shouldn't be e2e, just that nobody really had a choice to make it anything other than e2e. Now the follow-up question is --- as John Day pointed out, deployment difficulties have always been a fact of life, with the router/ISP companies being the primary reason back in the day. For me that raises the question: was e2e ever a clean philosophy, or was it simply the only practical deployment model open to designers twenty years back, as well? How many systems have been deployed end-to-end when people actually had a choice to do it some other way? And - do people have a tendency to use e2e the philosophy to incorrectly justify system designs that were e2e by necessity? (of course, sometimes the only deployment model turns out to be the correct one). (on the road currently, apologies in advance for late responses) - mahesh -----Original Message----- From: Rute Sofia [mailto:rsofia at inescporto.pt] Sent: Mon 4/21/2008 5:34 AM To: Christian Huitema Cc: Mahesh Balakrishnan; end2end-interest at postel.org Subject: Re: [e2e] end of interest Agree with Christian. And specifically focusing on L3, there seems to be other wave coming, with FON and now the more recent Whisher. New business roles (virtual ISPs, micro-ISPs) are also emerging. Rute Christian Huitema wrote: > In what world are you living, exactly? What about systems like Skype, or > Bit torrent? They are definitely pushing the envelope of end to end > designs, are widely deployed, and are not controlled by major corporations. > > > > *From:* end2end-interest-bounces at postel.org > [mailto:end2end-interest-bounces at postel.org] *On Behalf Of *Mahesh > Balakrishnan > *Sent:* Friday, April 18, 2008 5:11 PM > *To:* end2end-interest at postel.org > *Subject:* Re: [e2e] end of interest > > > > In fact there seems to be pushback from both ends --- we can't deploy > end-to-end protocols because major companies own the end-host stacks; > and we can't push mechanisms deep into the network because ISPs and > router companies own the network. Arguably the latter source of pushback > played a major role in the emergence of the e2e philosophy; but now we > have equally powerful commercial forces on the other side. > > So effectively the only practical mode of deployment seems to be the > 'almost' end-to-end middlebox --- one hop away from the end-host but not > quite into the network (and the Maelstrom work I presented day before > yesterday at NSDI would be one example). > > - mahesh > > -- > http://www.cs.cornell.edu/~mahesh > > > -----Original Message----- > From: end2end-interest-bounces at postel.org on behalf of David P. Reed > Sent: Fri 4/18/2008 9:13 AM > To: Jon Crowcroft > Cc: 'end2end-interest at postel.org' > Subject: Re: [e2e] end of interest > > I personally think that the network community has become frustrated with > the inability to explore end-to-end protocols because the endpoint > stacks are "locked in" by vendors in proprietary code. > > > -- Rute Sofia, PhD (rsofia at inescporto.pt, +351 22 2094263) http://www.inescporto.pt/utm/~ian/ Area Leader/Responsa'vel de A'rea IAN: Internet Architectures and Networking UTM: Telecommunications and Networking Unit INESC Porto -------------- next part -------------- An HTML attachment was scrubbed... URL: http://mailman.postel.org/pipermail/end2end-interest/attachments/20080421/619beeb6/attachment-0001.html From dpreed at reed.com Mon Apr 21 07:40:41 2008 From: dpreed at reed.com (David P. Reed) Date: Mon, 21 Apr 2008 10:40:41 -0400 Subject: [e2e] end of interest In-Reply-To: References: <48089E95.1090103@reed.com> <480908CC.5050400@reed.com> <1208556801.23430.2.camel@sioux.systems.cs.cornell.edu> <480A4483.8070703@reed.com> <12D20F05-721A-47B7-890E-C9DA143D9919@cisco.com> Message-ID: <480CA769.40607@reed.com> Jon Crowcroft wrote: > so the only solution is to kill off > routers and replace them entirely with middleboxes.... > > Seems like a good plan. Then we go after Akamai, take their CDN customers, and capture all the queries and build a better search platform, killing Google. I think Cisco would be silly not to push this as far as it can go. Not sure what role research plays in this, though. Seems like an MBA-heavy strategy. From kempf at docomolabs-usa.com Mon Apr 21 08:59:15 2008 From: kempf at docomolabs-usa.com (James Kempf) Date: Mon, 21 Apr 2008 08:59:15 -0700 Subject: [e2e] end of interest References: <48089E95.1090103@reed.com><48095BA4.3020709@cs.utk.edu><2656536D-323E-405A-AD55-B6F6653FCDFF@nokia.com> <480ABB18.1060305@isi.edu><6d2996bb0804200043i76986be9n32282fff2814ae02@mail.gmail.com> Message-ID: <026d01c8a3c8$a64278e0$1a6115ac@dcml.docomolabsusa.com> >the value of the net to users is that it connects them to content. the >network providers >are in the business of taking a fraction of the business that the content >providers are in >... If you look at any of the research on networks, most researchers agree that the value of the network is in connectivity. There's arguments about whether the value scales as O( n**2 ) via Metcalfe's Law or something more like O( n log(n) ) which Briscoe, Odlyzko, and Tilly claim. But nobody claims that the value of networks is in the bandwidth. Last time I looked, network providers weren't charging for connectivity, they were charging for bandwidth. Google makes tons of money off of small text ads that use almost no bandwidth but cash in from free connectivity. Network providers are forced to give away connectivity because the Internet architecture provides no way for them to charge for it. Not a particularly good business when you are forced to give away what is of value and charge for what isn't. My blog post this week discusses this more and the connection with end-to-end (http://cleanslate-internet.blogspot.com). jak From Jon.Crowcroft at cl.cam.ac.uk Mon Apr 21 09:10:27 2008 From: Jon.Crowcroft at cl.cam.ac.uk (Jon Crowcroft) Date: Mon, 21 Apr 2008 17:10:27 +0100 Subject: [e2e] end of interest In-Reply-To: <026d01c8a3c8$a64278e0$1a6115ac@dcml.docomolabsusa.com> References: <48089E95.1090103@reed.com><48095BA4.3020709@cs.utk.edu><2656536D-323E-405A-AD55-B6F6653FCDFF@nokia.com> <480ABB18.1060305@isi.edu><6d2996bb0804200043i76986be9n32282fff2814ae02@mail.gmail.com> <026d01c8a3c8$a64278e0$1a6115ac@dcml.docomolabsusa.com> Message-ID: so that leads to an interesting conclusion which might align business models with anti-spam and anti-ddos economic incentives. 1. charge a sender for the number of reachable recipients per unit time...doesn't hurt the average joe q public much, collects much money of supernodes, big server sites, and spammers/ddossers... 2. retire In missive <026d01c8a3c8$a64278e0$1a6115ac at dcml.docomolabsusa.com>, "J ames Kempf" typed: >>>the value of the net to users is that it connects them to content. the >>>network providers >>>are in the business of taking a fraction of the business that the content >>>providers are in >>>... >> >>If you look at any of the research on networks, most researchers agree that >>the value of the network is in connectivity. There's arguments about whether >>the value scales as O( n**2 ) via Metcalfe's Law or something more like >>O( n log(n) ) which Briscoe, Odlyzko, and Tilly claim. But nobody claims >>that the value of networks is in the bandwidth. >> >>Last time I looked, network providers weren't charging for connectivity, >>they were charging for bandwidth. Google makes tons of money off of small >>text ads that use almost no bandwidth but cash in from free connectivity. >>Network providers are forced to give away connectivity because the Internet >>architecture provides no way for them to charge for it. Not a particularly >>good business when you are forced to give away what is of value and charge >>for what isn't. >> >>My blog post this week discusses this more and the connection with >>end-to-end (http://cleanslate-internet.blogspot.com). >> >> jak >> >> cheers jon From david.borman at windriver.com Mon Apr 21 09:10:20 2008 From: david.borman at windriver.com (David Borman) Date: Mon, 21 Apr 2008 11:10:20 -0500 Subject: [e2e] SCTP popularity In-Reply-To: References: Message-ID: In general, as long as TCP or UDP provide the needed functionality, there is no motivation to switch to SCTP. SCTP is and will be used for applications that need the additional functionality that SCTP provides beyond what TCP or UDP can offer. For example, Carrier Grade Linux (CGL) is one place where SCTP is being used today. -David Borman On Apr 19, 2008, at 10:09 AM, Lars Eggert wrote: > On 2008-4-19, at 5:07, ext Alex Moura wrote: >> Inspired by the last thread, I'd like to know your opinions >> regarding SCTP >> adoption and general use. It will never happen, or it will follow >> the same >> process - and time - that IPv6 adoption is taking? > > Adoption - sure; as George has pointed out, IPFIX is using it, and > so do several other applications and application-layer protocols. > > General use as in by you and me behind our home NATs - not > currently. But the Linux firewall/NAT has SCTP and DCCP support (and > BSD has at least SCTP, I think), and if residential NAT vendors that > use those systems allow enabling that functionality, it may happen. > I grant you that that's a big if, because there is little incentive. > > Lars From kempf at docomolabs-usa.com Mon Apr 21 10:42:36 2008 From: kempf at docomolabs-usa.com (James Kempf) Date: Mon, 21 Apr 2008 10:42:36 -0700 Subject: [e2e] end of interest References: <48089E95.1090103@reed.com><48095BA4.3020709@cs.utk.edu><2656536D-323E-405A-AD55-B6F6653FCDFF@nokia.com> <480ABB18.1060305@isi.edu><6d2996bb0804200043i76986be9n32282fff2814ae02@mail.gmail.com> <026d01c8a3c8$a64278e0$1a6115ac@dcml.docomolabsusa.com> Message-ID: <031e01c8a3d7$167ac5a0$1a6115ac@dcml.docomolabsusa.com> Jon, According to Patrick Peterson at Ironport (presentation week before last the RSA Conference), the Storm worm made over 70,000 connections during the first 36 hours after infection. And the Storm network earns the spammers over $150 million in revenue through fufillment of Viagra orders through CandianPharmacy.com from Indian and Chinese sources. If the architecture were such that connectivity was not free, you can be sure that the ISPs, equipment vendors, and others in the ecosystem would do something to ensure that this didn't happen. Otherwise, infected users would be screaming about their bills. jak ----- Original Message ----- From: "Jon Crowcroft" To: "James Kempf" Cc: "Saikat Guha" ; Sent: Monday, April 21, 2008 9:10 AM Subject: Re: [e2e] end of interest so that leads to an interesting conclusion which might align business models with anti-spam and anti-ddos economic incentives. 1. charge a sender for the number of reachable recipients per unit time...doesn't hurt the average joe q public much, collects much money of supernodes, big server sites, and spammers/ddossers... 2. retire In missive <026d01c8a3c8$a64278e0$1a6115ac at dcml.docomolabsusa.com>, "J ames Kempf" typed: >>>the value of the net to users is that it connects them to content. the >>>network providers >>>are in the business of taking a fraction of the business that the content >>>providers are in >>>... >> >>If you look at any of the research on networks, most researchers agree that >>the value of the network is in connectivity. There's arguments about whether >>the value scales as O( n**2 ) via Metcalfe's Law or something more like >>O( n log(n) ) which Briscoe, Odlyzko, and Tilly claim. But nobody claims >>that the value of networks is in the bandwidth. >> >>Last time I looked, network providers weren't charging for connectivity, >>they were charging for bandwidth. Google makes tons of money off of small >>text ads that use almost no bandwidth but cash in from free connectivity. >>Network providers are forced to give away connectivity because the Internet >>architecture provides no way for them to charge for it. Not a particularly >>good business when you are forced to give away what is of value and charge >>for what isn't. >> >>My blog post this week discusses this more and the connection with >>end-to-end (http://cleanslate-internet.blogspot.com). >> >> jak >> >> cheers jon From aaa at cs.stanford.edu Mon Apr 21 11:19:22 2008 From: aaa at cs.stanford.edu (Amr A. Awadallah) Date: Mon, 21 Apr 2008 11:19:22 -0700 Subject: [e2e] end of interest In-Reply-To: <026d01c8a3c8$a64278e0$1a6115ac@dcml.docomolabsusa.com> References: <48089E95.1090103@reed.com><48095BA4.3020709@cs.utk.edu><2656536D-323E-405A-AD55-B6F6653FCDFF@nokia.com> <480ABB18.1060305@isi.edu><6d2996bb0804200043i76986be9n32282fff2814ae02@mail.gmail.com> <026d01c8a3c8$a64278e0$1a6115ac@dcml.docomolabsusa.com> Message-ID: <480CDAAA.8070701@cs.stanford.edu> > My blog post this week discusses this more and the connection with end-to-end (http://cleanslate-internet.blogspot.com this is the reason why e2e (as a mailing list) is dieing, we are just reading/commenting to our blogs instead :) -- amr James Kempf wrote: >> the value of the net to users is that it connects them to content. >> the network providers >> are in the business of taking a fraction of the business that the >> content providers are in >> ... > > If you look at any of the research on networks, most researchers agree > that the value of the network is in connectivity. There's arguments > about whether the value scales as O( n**2 ) via Metcalfe's Law or > something more like O( n log(n) ) which Briscoe, Odlyzko, and Tilly > claim. But nobody claims that the value of networks is in the bandwidth. > > Last time I looked, network providers weren't charging for > connectivity, they were charging for bandwidth. Google makes tons of > money off of small text ads that use almost no bandwidth but cash in > from free connectivity. Network providers are forced to give away > connectivity because the Internet architecture provides no way for > them to charge for it. Not a particularly good business when you are > forced to give away what is of value and charge for what isn't. > > My blog post this week discusses this more and the connection with > end-to-end (http://cleanslate-internet.blogspot.com). > > jak > From mellia at tlc.polito.it Mon Apr 21 11:47:23 2008 From: mellia at tlc.polito.it (Marco Mellia) Date: Mon, 21 Apr 2008 20:47:23 +0200 Subject: [e2e] end of interest In-Reply-To: <480ABC7B.8040208@isi.edu> References: <480ABC7B.8040208@isi.edu> Message-ID: <480CE13B.5050807@tlc.polito.it> Ok, so to avoid that April will have NO CFP, here is one :) ------------------------------ [Our apologies if you receive multiple copies of this CFP] Deadline is fast approaching. ================================================================= Call for Papers Special Issue of Computer Networks on Traffic classification and its applications to modern networks http://www.elsevier.com/authored_subject_sections/P05/CFP/cfp_trafficclass.pdf ----------------- Important Dates ------------------------- Manuscripts due: May 1, 2008 Notification to authors: September 10, 2008 Revised papers due: October 15, 2008 Accepted, revised papers to production: October 30, 2008 Publication: late 2008/early 2009 ----------------- Guest Editors ------------------------------ Marco Mellia, Politecnico di Torino, mellia at tlc.polito.it Antonio Pescape', Universita' di Napoli Federico II, pescape at unina.it Luca Salgarelli, Universit? di Brescia, luca.salgarelli at ing.unibs.it ================================================================= Ciao, /\/\/\rco +-----------------------------------+ | Marco Mellia - Assistant Professor| | Skypeid: mgmellia | | Tel: +39-011-564-4173 | | Cel: +39-340-9674888 | /"\ .. . . . . . . . . . . . . | Politecnico di Torino | \ / . ASCII Ribbon Campaign . | Corso Duca degli Abruzzi 24 | X .- NO HTML/RTF in e-mail . | Torino - 10129 - Italy | / \ .- NO Word docs in e-mail. | http://www.telematica.polito.it | .. . . . . . . . . . . . . +-----------------------------------+ The box said "Requires Windows 95 or Better." So I installed Linux. > > Jon Crowcroft wrote: >> last month's postings on this list was 80% CfPs; >> this month has 1 message which is a CfP; >> is this the end of interest in end-to-end? > > Just a note from your list admin: > > True, March was 80% CFPs. > > Feb was only 20%, and Jan was 3 out of 225 (1.33%). Overall, it just > seems more like people were handling midterms, off on spring break, or > otherwise temporarily occupied. > > I'll wait for that trend to persist for more than a month before I > call for the paramedics, though discussion triggers are always welcome... > > Joe > From dpreed at reed.com Mon Apr 21 12:02:26 2008 From: dpreed at reed.com (David P. Reed) Date: Mon, 21 Apr 2008 15:02:26 -0400 Subject: [e2e] end of interest In-Reply-To: <031e01c8a3d7$167ac5a0$1a6115ac@dcml.docomolabsusa.com> References: <48089E95.1090103@reed.com><48095BA4.3020709@cs.utk.edu><2656536D-323E-405A-AD55-B6F6653FCDFF@nokia.com> <480ABB18.1060305@isi.edu><6d2996bb0804200043i76986be9n32282fff2814ae02@mail.gmail.com> <026d01c8a3c8$a64278e0$1a6115ac@dcml.docomolabsusa.com> <031e01c8a3d7$167ac5a0$1a6115ac@dcml.docomolabsusa.com> Message-ID: <480CE4C2.2060501@reed.com> James - Connectivity in the Internet is NOT free (at least "as in beer"). The Storm worm used already paid-up capacity, paid for by the users in bulk form. One can get confused because the phone companies charge "by the minute" for some things, but give away *for free* so-to-speak the always-on connection between home instrument and line card. How dare you not pay for receiving all the bits of information that come from knowing that "no one is calling you right now" - you should pay at texting rates for those bits, unless you are a pirate. :-) I'd suggest using accurate terminology, lest one just add to the confusion. The primary cost base of a communications company is composed of capex for the wires and switches, power for them, and salaries for humans who interface with customers. None of those are correlated even slightly with minutes of voice or bits of data. Economics is hard work, but its probably worth learning the meaning of the term "free". James Kempf wrote: > Jon, > > According to Patrick Peterson at Ironport (presentation week before > last the RSA Conference), the Storm worm made over 70,000 connections > during the first 36 hours after infection. And the Storm network earns > the spammers over $150 million in revenue through fufillment of Viagra > orders through CandianPharmacy.com from Indian and Chinese sources. > > If the architecture were such that connectivity was not free, you can > be sure that the ISPs, equipment vendors, and others in the ecosystem > would do something to ensure that this didn't happen. Otherwise, > infected users would be screaming about their bills. > > jak > > ----- Original Message ----- From: "Jon Crowcroft" > > To: "James Kempf" > Cc: "Saikat Guha" ; > Sent: Monday, April 21, 2008 9:10 AM > Subject: Re: [e2e] end of interest > > > so that leads to an interesting conclusion which might align > business models with anti-spam and anti-ddos economic > incentives. > > 1. charge a sender for the number of reachable recipients > per unit time...doesn't hurt the average joe q public much, > collects much money of supernodes, big server sites, and > spammers/ddossers... > > 2. retire > > > In missive <026d01c8a3c8$a64278e0$1a6115ac at dcml.docomolabsusa.com>, "J > ames Kempf" typed: > > >>>the value of the net to users is that it connects them to content. the > >>>network providers > >>>are in the business of taking a fraction of the business that the > content > >>>providers are in > >>>... > >> > >>If you look at any of the research on networks, most researchers > agree that > >>the value of the network is in connectivity. There's arguments about > whether > >>the value scales as O( n**2 ) via Metcalfe's Law or something more > like > >>O( n log(n) ) which Briscoe, Odlyzko, and Tilly claim. But nobody > claims > >>that the value of networks is in the bandwidth. > >> > >>Last time I looked, network providers weren't charging for > connectivity, > >>they were charging for bandwidth. Google makes tons of money off of > small > >>text ads that use almost no bandwidth but cash in from free > connectivity. > >>Network providers are forced to give away connectivity because the > Internet > >>architecture provides no way for them to charge for it. Not a > particularly > >>good business when you are forced to give away what is of value and > charge > >>for what isn't. > >> > >>My blog post this week discusses this more and the connection with > >>end-to-end (http://cleanslate-internet.blogspot.com). > >> > >> jak > >> > >> > > cheers > > jon > > > > From kempf at docomolabs-usa.com Mon Apr 21 12:28:33 2008 From: kempf at docomolabs-usa.com (James Kempf) Date: Mon, 21 Apr 2008 12:28:33 -0700 Subject: [e2e] end of interest References: <48089E95.1090103@reed.com><48095BA4.3020709@cs.utk.edu><2656536D-323E-405A-AD55-B6F6653FCDFF@nokia.com> <480ABB18.1060305@isi.edu><6d2996bb0804200043i76986be9n32282fff2814ae02@mail.gmail.com> <026d01c8a3c8$a64278e0$1a6115ac@dcml.docomolabsusa.com> <480CDAAA.8070701@cs.stanford.edu> Message-ID: <041e01c8a3e5$e30b8380$1a6115ac@dcml.docomolabsusa.com> Amr, Actually, if you know of any blogs that are discussing this kind of stuff, I'd love to hear about them. I did a Google search for blogs on clean slate Internet design and it came up negative, which is why I started mine. Most of the blogs in network technology areas I've been able to find are either on security related topics (like Bruce Schneier's excellent blog) or reviews of new gadgets and breathless anticipation of Apple's next iPhone release. jak ----- Original Message ----- From: "Amr A. Awadallah" To: "James Kempf" Cc: "Jon Crowcroft" ; "Saikat Guha" ; Sent: Monday, April 21, 2008 11:19 AM Subject: Re: [e2e] end of interest > My blog post this week discusses this more and the connection with end-to-end (http://cleanslate-internet.blogspot.com this is the reason why e2e (as a mailing list) is dieing, we are just reading/commenting to our blogs instead :) -- amr James Kempf wrote: >> the value of the net to users is that it connects them to content. >> the network providers >> are in the business of taking a fraction of the business that the >> content providers are in >> ... > > If you look at any of the research on networks, most researchers agree > that the value of the network is in connectivity. There's arguments > about whether the value scales as O( n**2 ) via Metcalfe's Law or > something more like O( n log(n) ) which Briscoe, Odlyzko, and Tilly > claim. But nobody claims that the value of networks is in the bandwidth. > > Last time I looked, network providers weren't charging for > connectivity, they were charging for bandwidth. Google makes tons of > money off of small text ads that use almost no bandwidth but cash in > from free connectivity. Network providers are forced to give away > connectivity because the Internet architecture provides no way for > them to charge for it. Not a particularly good business when you are > forced to give away what is of value and charge for what isn't. > > My blog post this week discusses this more and the connection with > end-to-end (http://cleanslate-internet.blogspot.com). > > jak > From kempf at docomolabs-usa.com Mon Apr 21 12:47:08 2008 From: kempf at docomolabs-usa.com (James Kempf) Date: Mon, 21 Apr 2008 12:47:08 -0700 Subject: [e2e] end of interest References: <48089E95.1090103@reed.com><48095BA4.3020709@cs.utk.edu><2656536D-323E-405A-AD55-B6F6653FCDFF@nokia.com> <480ABB18.1060305@isi.edu><6d2996bb0804200043i76986be9n32282fff2814ae02@mail.gmail.com> <026d01c8a3c8$a64278e0$1a6115ac@dcml.docomolabsusa.com> <031e01c8a3d7$167ac5a0$1a6115ac@dcml.docomolabsusa.com> <480CE4C2.2060501@reed.com> Message-ID: <044c01c8a3e8$7bbea150$1a6115ac@dcml.docomolabsusa.com> David, When I say "for free", what I meant was that when I get my bill, what does it say I'm paying for? What it says is that I am paying $25 a month for 1 Mpbs downlink and 768 kbps uplink for the month of, say, March (AT&T just raised my rates). It does not say that I am paying $x for 250 connections, above which I would have to pay incrementally for additional connections (and pay through the nose if I get a Storm worm infection). To see what a difference this makes, consider the California "energy crisis" of the early 2000's. Utilities were required to buy power at market rate, but only allowed to charge customers a fixed and regulated price. Enron took advantage of this situation (kind of like the Storm work authors and CanadianPharmacy.com do with the Internet) to make enormous amounts of money by manipulating the market. Getting pricing signals right is very important because they influence customer and market participant behavior. Getting them wrong can lead to all kinds of distortions. If, as the theoretical studies suggest, the value of networks is in their connectivity, then why shouldn't businesses charge for that value? Or are you suggesting that the theoretical studies on network value have it wrong? Your point about capex of equipment and labor to maintain the network has nothing to do with the value provided by the network, it has to do with the cost of providing the service. These are two different things. jak ----- Original Message ----- From: "David P. Reed" To: "James Kempf" Cc: "Jon Crowcroft" ; Sent: Monday, April 21, 2008 12:02 PM Subject: Re: [e2e] end of interest James - Connectivity in the Internet is NOT free (at least "as in beer"). The Storm worm used already paid-up capacity, paid for by the users in bulk form. One can get confused because the phone companies charge "by the minute" for some things, but give away *for free* so-to-speak the always-on connection between home instrument and line card. How dare you not pay for receiving all the bits of information that come from knowing that "no one is calling you right now" - you should pay at texting rates for those bits, unless you are a pirate. :-) I'd suggest using accurate terminology, lest one just add to the confusion. The primary cost base of a communications company is composed of capex for the wires and switches, power for them, and salaries for humans who interface with customers. None of those are correlated even slightly with minutes of voice or bits of data. Economics is hard work, but its probably worth learning the meaning of the term "free". James Kempf wrote: > Jon, > > According to Patrick Peterson at Ironport (presentation week before > last the RSA Conference), the Storm worm made over 70,000 connections > during the first 36 hours after infection. And the Storm network earns > the spammers over $150 million in revenue through fufillment of Viagra > orders through CandianPharmacy.com from Indian and Chinese sources. > > If the architecture were such that connectivity was not free, you can > be sure that the ISPs, equipment vendors, and others in the ecosystem > would do something to ensure that this didn't happen. Otherwise, > infected users would be screaming about their bills. > > jak > > ----- Original Message ----- From: "Jon Crowcroft" > > To: "James Kempf" > Cc: "Saikat Guha" ; > Sent: Monday, April 21, 2008 9:10 AM > Subject: Re: [e2e] end of interest > > > so that leads to an interesting conclusion which might align > business models with anti-spam and anti-ddos economic > incentives. > > 1. charge a sender for the number of reachable recipients > per unit time...doesn't hurt the average joe q public much, > collects much money of supernodes, big server sites, and > spammers/ddossers... > > 2. retire > > > In missive <026d01c8a3c8$a64278e0$1a6115ac at dcml.docomolabsusa.com>, "J > ames Kempf" typed: > > >>>the value of the net to users is that it connects them to content. the > >>>network providers > >>>are in the business of taking a fraction of the business that the > content > >>>providers are in > >>>... > >> > >>If you look at any of the research on networks, most researchers > agree that > >>the value of the network is in connectivity. There's arguments about > whether > >>the value scales as O( n**2 ) via Metcalfe's Law or something more > like > >>O( n log(n) ) which Briscoe, Odlyzko, and Tilly claim. But nobody > claims > >>that the value of networks is in the bandwidth. > >> > >>Last time I looked, network providers weren't charging for > connectivity, > >>they were charging for bandwidth. Google makes tons of money off of > small > >>text ads that use almost no bandwidth but cash in from free > connectivity. > >>Network providers are forced to give away connectivity because the > Internet > >>architecture provides no way for them to charge for it. Not a > particularly > >>good business when you are forced to give away what is of value and > charge > >>for what isn't. > >> > >>My blog post this week discusses this more and the connection with > >>end-to-end (http://cleanslate-internet.blogspot.com). > >> > >> jak > >> > >> > > cheers > > jon > > > > From Jon.Crowcroft at cl.cam.ac.uk Mon Apr 21 12:52:26 2008 From: Jon.Crowcroft at cl.cam.ac.uk (Jon Crowcroft) Date: Mon, 21 Apr 2008 20:52:26 +0100 Subject: [e2e] end of interest In-Reply-To: <044c01c8a3e8$7bbea150$1a6115ac@dcml.docomolabsusa.com> References: <48089E95.1090103@reed.com><48095BA4.3020709@cs.utk.edu><2656536D-323E-405A-AD55-B6F6653FCDFF@nokia.com> <480ABB18.1060305@isi.edu><6d2996bb0804200043i76986be9n32282fff2814ae02@mail.gmail.com> <026d01c8a3c8$a64278e0$1a6115ac@dcml.docomolabsusa.com> <031e01c8a3d7$167ac5a0$1a6115ac@dcml.docomolabsusa.com> <480CE4C2.2060501@reed.com> <044c01c8a3e8$7bbea150$1a6115ac@dcml.docomolabsusa.com> Message-ID: so the value of the net is the value the end users bring when they contribute content - this (as you rightly say) is somewhere in the n..n*ln(0n) ..n^2 range... so the problem is that any given ISP cannot bring more than a fraction of this - but it is also not at all clear that the peering v. customer/provider model, nor a simple count of direct subscribers, would really tell you either... what you need is pagerank for ISPs... and then a charge to the end user based on the so called "value" of the set of places they visited... only probem is that the direction of value is completely obscured by the compelx arrangements content providers make to send you the stuff you want (should pay for) and the stuff you dont want (adverts to get them revenue so they dont have to charge you for the stuff you want). this applies even to personal sites (like i can get adsense revenue on my blogs)... so which end pays how much? hmmm... ...no, i give up- i think the state should own and run the net - there's clearly no long term value in it with the way it is architected today.. In missive <044c01c8a3e8$7bbea150$1a6115ac at dcml.docomolabsusa.com>, "James Kemp f" typed: >>David, >> >>When I say "for free", what I meant was that when I get my bill, what does >>it say I'm paying for? >> >>What it says is that I am paying $25 a month for 1 Mpbs downlink and 768 >>kbps uplink for the month of, say, March (AT&T just raised my rates). >> >>It does not say that I am paying $x for 250 connections, above which I would >>have to pay incrementally for additional connections (and pay through the >>nose if I get a Storm worm infection). >> >>To see what a difference this makes, consider the California "energy crisis" >>of the early 2000's. Utilities were required to buy power at market rate, >>but only allowed to charge customers a fixed and regulated price. Enron took >>advantage of this situation (kind of like the Storm work authors and >>CanadianPharmacy.com do with the Internet) to make enormous amounts of money >>by manipulating the market. Getting pricing signals right is very important >>because they influence customer and market participant behavior. Getting >>them wrong can lead to all kinds of distortions. >> >>If, as the theoretical studies suggest, the value of networks is in their >>connectivity, then why shouldn't businesses charge for that value? Or are >>you suggesting that the theoretical studies on network value have it wrong? >>Your point about capex of equipment and labor to maintain the network has >>nothing to do with the value provided by the network, it has to do with the >>cost of providing the service. These are two different things. >> >> jak >> >>----- Original Message ----- >>From: "David P. Reed" >>To: "James Kempf" >>Cc: "Jon Crowcroft" ; >> >>Sent: Monday, April 21, 2008 12:02 PM >>Subject: Re: [e2e] end of interest >> >> >>James - >> >>Connectivity in the Internet is NOT free (at least "as in beer"). The >>Storm worm used already paid-up capacity, paid for by the users in bulk >>form. One can get confused because the phone companies charge "by the >>minute" for some things, but give away *for free* so-to-speak the >>always-on connection between home instrument and line card. How dare >>you not pay for receiving all the bits of information that come from >>knowing that "no one is calling you right now" - you should pay at >>texting rates for those bits, unless you are a pirate. :-) >> >>I'd suggest using accurate terminology, lest one just add to the confusion. >> >>The primary cost base of a communications company is composed of capex >>for the wires and switches, power for them, and salaries for humans who >>interface with customers. None of those are correlated even slightly >>with minutes of voice or bits of data. >> >>Economics is hard work, but its probably worth learning the meaning of >>the term "free". >> >> >>James Kempf wrote: >>> Jon, >>> >>> According to Patrick Peterson at Ironport (presentation week before >>> last the RSA Conference), the Storm worm made over 70,000 connections >>> during the first 36 hours after infection. And the Storm network earns >>> the spammers over $150 million in revenue through fufillment of Viagra >>> orders through CandianPharmacy.com from Indian and Chinese sources. >>> >>> If the architecture were such that connectivity was not free, you can >>> be sure that the ISPs, equipment vendors, and others in the ecosystem >>> would do something to ensure that this didn't happen. Otherwise, >>> infected users would be screaming about their bills. >>> >>> jak >>> >>> ----- Original Message ----- From: "Jon Crowcroft" >>> >>> To: "James Kempf" >>> Cc: "Saikat Guha" ; >>> Sent: Monday, April 21, 2008 9:10 AM >>> Subject: Re: [e2e] end of interest >>> >>> >>> so that leads to an interesting conclusion which might align >>> business models with anti-spam and anti-ddos economic >>> incentives. >>> >>> 1. charge a sender for the number of reachable recipients >>> per unit time...doesn't hurt the average joe q public much, >>> collects much money of supernodes, big server sites, and >>> spammers/ddossers... >>> >>> 2. retire >>> >>> >>> In missive <026d01c8a3c8$a64278e0$1a6115ac at dcml.docomolabsusa.com>, "J >>> ames Kempf" typed: >>> >>> >>>the value of the net to users is that it connects them to content. the >>> >>>network providers >>> >>>are in the business of taking a fraction of the business that the >>> content >>> >>>providers are in >>> >>>... >>> >> >>> >>If you look at any of the research on networks, most researchers >>> agree that >>> >>the value of the network is in connectivity. There's arguments about >>> whether >>> >>the value scales as O( n**2 ) via Metcalfe's Law or something more >>> like >>> >>O( n log(n) ) which Briscoe, Odlyzko, and Tilly claim. But nobody >>> claims >>> >>that the value of networks is in the bandwidth. >>> >> >>> >>Last time I looked, network providers weren't charging for >>> connectivity, >>> >>they were charging for bandwidth. Google makes tons of money off of >>> small >>> >>text ads that use almost no bandwidth but cash in from free >>> connectivity. >>> >>Network providers are forced to give away connectivity because the >>> Internet >>> >>architecture provides no way for them to charge for it. Not a >>> particularly >>> >>good business when you are forced to give away what is of value and >>> charge >>> >>for what isn't. >>> >> >>> >>My blog post this week discusses this more and the connection with >>> >>end-to-end (http://cleanslate-internet.blogspot.com). >>> >> >>> >> jak >>> >> >>> >> >>> >>> cheers >>> >>> jon >>> >>> >>> >>> >> >> cheers jon From dpreed at reed.com Mon Apr 21 13:29:15 2008 From: dpreed at reed.com (David P. Reed) Date: Mon, 21 Apr 2008 16:29:15 -0400 Subject: [e2e] end of interest In-Reply-To: <044c01c8a3e8$7bbea150$1a6115ac@dcml.docomolabsusa.com> References: <48089E95.1090103@reed.com><48095BA4.3020709@cs.utk.edu><2656536D-323E-405A-AD55-B6F6653FCDFF@nokia.com> <480ABB18.1060305@isi.edu><6d2996bb0804200043i76986be9n32282fff2814ae02@mail.gmail.com> <026d01c8a3c8$a64278e0$1a6115ac@dcml.docomolabsusa.com> <031e01c8a3d7$167ac5a0$1a6115ac@dcml.docomolabsusa.com> <480CE4C2.2060501@reed.com> <044c01c8a3e8$7bbea150$1a6115ac@dcml.docomolabsusa.com> Message-ID: <480CF91B.4010909@reed.com> James Kempf wrote: > David, > > When I say "for free", what I meant was that when I get my bill, what > does it say I'm paying for? > Mine says 1 month of "Internet service" - whatever that is. :-) But seriously, I agree that we often don't talk about the bill as including "options to connect" in pairs, groups, etc. that the *switching* aspect of Internet service affords us. In finance, option value inherent in purchasing a commodity contract is understood, so we ought to presume that option value related to switching is understood as bundled with the purchase of commodity Internet service. Obviously I think that switching and routing is valuable or I wouldn't have written about the economics of group-forming networks, and I wouldn't have agreed with Metcalfe that the N^2 law measures a form of utility that is separate from mere fixed-line bitrate. However, I think when we customers buy Internet service (or for that matter, cable TV service) we understand at least implicitly we are not paying for a flood of bits alone. The value of flexibility to choose what bits we want to send and receive, and where we want to send and receive them at any instant of time is something a service subscriber expects. From touch at ISI.EDU Tue Apr 22 07:25:16 2008 From: touch at ISI.EDU (Joe Touch) Date: Tue, 22 Apr 2008 07:25:16 -0700 Subject: [e2e] end of interest In-Reply-To: <480CE13B.5050807@tlc.polito.it> References: <480ABC7B.8040208@isi.edu> <480CE13B.5050807@tlc.polito.it> Message-ID: <480DF54C.6090200@isi.edu> REMINDER to all: CFPs to this list must be approved in advance. Joe (as list admin) Marco Mellia wrote: > Ok, so to avoid that April will have NO CFP, here is one :) > ------------------------------ > [Our apologies if you receive multiple copies of this CFP] > > Deadline is fast approaching. > ================================================================= > > Call for Papers > > Special Issue of Computer Networks on Traffic classification and its > applications to modern networks > > http://www.elsevier.com/authored_subject_sections/P05/CFP/cfp_trafficclass.pdf > > > ----------------- Important Dates ------------------------- > Manuscripts due: May 1, 2008 > Notification to authors: September 10, 2008 > Revised papers due: October 15, 2008 > Accepted, revised papers to production: October 30, 2008 > Publication: late 2008/early 2009 > > ----------------- Guest Editors ------------------------------ > Marco Mellia, Politecnico di Torino, mellia at tlc.polito.it > Antonio Pescape', Universita' di Napoli Federico II, pescape at unina.it > Luca Salgarelli, Universit? di Brescia, luca.salgarelli at ing.unibs.it > > ================================================================= > > Ciao, /\/\/\rco > > +-----------------------------------+ > | Marco Mellia - Assistant Professor| > | Skypeid: mgmellia | > | Tel: +39-011-564-4173 | > | Cel: +39-340-9674888 | /"\ .. . . . . . . . . . . . . > | Politecnico di Torino | \ / . ASCII Ribbon Campaign . > | Corso Duca degli Abruzzi 24 | X .- NO HTML/RTF in e-mail . > | Torino - 10129 - Italy | / \ .- NO Word docs in e-mail. > | http://www.telematica.polito.it | .. . . . . . . . . . . . . > +-----------------------------------+ > The box said "Requires Windows 95 or Better." So I installed Linux. > > >> >> Jon Crowcroft wrote: >>> last month's postings on this list was 80% CfPs; >>> this month has 1 message which is a CfP; >>> is this the end of interest in end-to-end? >> >> Just a note from your list admin: >> >> True, March was 80% CFPs. >> >> Feb was only 20%, and Jan was 3 out of 225 (1.33%). Overall, it just >> seems more like people were handling midterms, off on spring break, or >> otherwise temporarily occupied. >> >> I'll wait for that trend to persist for more than a month before I >> call for the paramedics, though discussion triggers are always welcome... >> >> Joe >> -------------- next part -------------- A non-text attachment was scrubbed... Name: signature.asc Type: application/pgp-signature Size: 250 bytes Desc: OpenPGP digital signature Url : http://mailman.postel.org/pipermail/end2end-interest/attachments/20080422/6bb3f12d/signature.bin From dhc2 at dcrocker.net Tue Apr 22 20:55:08 2008 From: dhc2 at dcrocker.net (Dave Crocker) Date: Tue, 22 Apr 2008 20:55:08 -0700 Subject: [e2e] experimenting on customers In-Reply-To: References: <48089E95.1090103@reed.com> <480908CC.5050400@reed.com> <480C9C37.40509@reed.com> Message-ID: <480EB31C.7060403@dcrocker.net> Jon Crowcroft wrote: > I dont understand all this - most software in the last 30 > years is an experiment on customers - the internet as a whole > is an experiment ... > so if we are honest, we'd admit this and say > what we need is a pharma model of informed consent > yeah, even discounts In looking over the many postings in this thread, the word "experiment" provides the most leverage both for insight and for confusion. Experiments come in very different flavors, notably with very different risk. When talking about something on the scale of the current, public Internet, or American democracy or global jet travel, the term "experiment" reminds us that we do not fully understand impact. But the term also denotes a risk of failure which cannot reasonably apply for these grander uses. (After a few hundred years, if a civilization dies off, is it a "failure", even though we label it an experiment?) In other words, we use the word "experiment" here in a non-technical way, connoting the unknown, rather the denoting controlled manipulation, diligent study and incremental refinement. So, some of the complaints about being unable to experiment on the open Internet simply do not make sense, any more than "testing" a radically new concrete -- with no use experience -- on a freeway bridge would make sense. Risk is obviously too high; in fact, failure early in the lifecycle of a new technology is typically guaranteed. Would you drive over that sucker? Or under it? So if someone is going to express concerns about barriers to adoption, such as a lack of flexibility by providers or product companies, they need to accompany it will a compelling adoption case that shows sufficiently low risk and sufficiently high benefit. Typically, that needs to come from real experimentation, meaning early-stage development, real testing, and pilot deployment. (Quite nicely this has the not-so-minor side benefit of grooming an increasingly significant constituency that wants the technology adopted.) Businesses do not deploy real experiments in their products and services. Costs and risks are far both too high. What they deploy are features that provide relatively assured benefits. As for "blocking" experiments by others, think again of the bridge. Collateral damage requires that public infrastructure services been particularly conservative in permitting change. In the early 90s, when new routing protocols were being defined and debated, it was also noted that there was no 'laboratory' large enough to test the protocols to scale, prior to deployment in the open Internet. One thought was to couple smaller test networks, via tunnels across the Internet. I suppose Internet II counts as a modern alternative. In other words, real experiments needs real laboratories. The other challenge for this particular thread is that the term end-to-end is treated as a rigid absolute, but never has actually been that. It is a term of relativity defined by two boundary points. The the modern Internet has added more complexity between the points, as others have noted. Rather than a simplistic host-net dichotomy we have layers of intermediary nets, and often layers of application hosts. (Thank you, akamai...) We also have layers outside of what we typically call end-points, such as services that treat email as underlying infrastructure, rather than "the" application. And we have layers of trust (tussles). And, and, and... So when claiming to need end-to-end, the question is which of many possible ends? And, for what purpose, or one might say... to what end? d/ -- Dave Crocker Brandenburg InternetWorking bbiw.net From dpreed at reed.com Wed Apr 23 05:19:59 2008 From: dpreed at reed.com (David P. Reed) Date: Wed, 23 Apr 2008 08:19:59 -0400 Subject: [e2e] experimenting on customers In-Reply-To: <480EB31C.7060403@dcrocker.net> References: <48089E95.1090103@reed.com> <480908CC.5050400@reed.com> <480C9C37.40509@reed.com> <480EB31C.7060403@dcrocker.net> Message-ID: <480F296F.7020807@reed.com> Dave - as I mentioned earlier, there is a huge difference between experimenting on customers and letting customers experiment. Your post equates the two. I suggest that the distinction is crucial. And that is the point of the end-to-end argument, at least 80% of it, when applied to modularity between carriers and users, or the modularity between systems vendors and users, or the modularity between companies that would hope to support innovative research and researchers. Dave Crocker wrote: > > > Jon Crowcroft wrote: >> I dont understand all this - most software in the last 30 >> years is an experiment on customers - the internet as a whole >> is an experiment > ... >> so if we are honest, we'd admit this and say >> what we need is a pharma model of informed consent >> yeah, even discounts > > > In looking over the many postings in this thread, the word > "experiment" provides the most leverage both for insight and for > confusion. > > Experiments come in very different flavors, notably with very > different risk. > When talking about something on the scale of the current, public > Internet, > or American democracy or global jet travel, the term "experiment" > reminds us > that we do not fully understand impact. But the term also denotes a > risk of > failure which cannot reasonably apply for these grander uses. (After > a few > hundred years, if a civilization dies off, is it a "failure", even > though we > label it an experiment?) In other words, we use the word "experiment" > here in > a non-technical way, connoting the unknown, rather the denoting > controlled > manipulation, diligent study and incremental refinement. > > So, some of the complaints about being unable to experiment on the open > Internet simply do not make sense, any more than "testing" a radically > new concrete -- with no use experience -- on a freeway bridge would > make sense. Risk is obviously too high; in fact, failure early in the > lifecycle of a new > technology is typically guaranteed. Would you drive over that sucker? > Or under it? > > So if someone is going to express concerns about barriers to adoption, > such as > a lack of flexibility by providers or product companies, they need to > accompany it will a compelling adoption case that shows sufficiently > low risk > and sufficiently high benefit. Typically, that needs to come from real > experimentation, meaning early-stage development, real testing, and pilot > deployment. (Quite nicely this has the not-so-minor side benefit of > grooming an increasingly significant constituency that wants the > technology adopted.) > > Businesses do not deploy real experiments in their products and services. > Costs and risks are far both too high. What they deploy are features > that provide relatively assured benefits. > > As for "blocking" experiments by others, think again of the bridge. > Collateral damage requires that public infrastructure services been > particularly conservative in permitting change. > > In the early 90s, when new routing protocols were being defined and > debated, it was also noted that there was no 'laboratory' large enough > to test the protocols to scale, prior to deployment in the open > Internet. One thought was to couple smaller test networks, via > tunnels across the Internet. I suppose Internet II counts as a modern > alternative. In other words, real experiments needs real laboratories. > > The other challenge for this particular thread is that the term > end-to-end is > treated as a rigid absolute, but never has actually been that. It is > a term > of relativity defined by two boundary points. The the modern Internet > has added more complexity between the points, as others have noted. > Rather than a simplistic host-net dichotomy we have layers of > intermediary nets, and often layers of application hosts. (Thank you, > akamai...) We also have layers outside of what we typically call > end-points, such as services that treat email as underlying > infrastructure, rather than "the" application. > > And we have layers of trust (tussles). > > And, and, and... > > So when claiming to need end-to-end, the question is which of many > possible ends? > > And, for what purpose, or one might say... to what end? > > d/ > > > > > > > > > > From dedutta at cisco.com Mon Apr 21 13:27:53 2008 From: dedutta at cisco.com (Debo Dutta (dedutta)) Date: Mon, 21 Apr 2008 13:27:53 -0700 Subject: [e2e] end of interest In-Reply-To: <480CDAAA.8070701@cs.stanford.edu> References: <48089E95.1090103@reed.com><48095BA4.3020709@cs.utk.edu><2656536D-323E-405A-AD55-B6F6653FCDFF@nokia.com> <480ABB18.1060305@isi.edu><6d2996bb0804200043i76986be9n32282fff2814ae02@mail.gmail.com> <026d01c8a3c8$a64278e0$1a6115ac@dcml.docomolabsusa.com> <480CDAAA.8070701@cs.stanford.edu> Message-ID: <0079EA505DB7BF48BE3D565D16ED55F80660E83F@xmb-sjc-225.amer.cisco.com> Maybe we should have a e2e blog and/or a e2e wiki with write access to all e2e members :) Or have a e2e group on facebook ..... :) Debo -----Original Message----- From: end2end-interest-bounces at postel.org [mailto:end2end-interest-bounces at postel.org] On Behalf Of Amr A. Awadallah Sent: Monday, April 21, 2008 11:19 AM To: James Kempf Cc: Jon Crowcroft; end2end-interest at postel.org Subject: Re: [e2e] end of interest > My blog post this week discusses this more and the connection with end-to-end (http://cleanslate-internet.blogspot.com this is the reason why e2e (as a mailing list) is dieing, we are just reading/commenting to our blogs instead :) -- amr James Kempf wrote: >> the value of the net to users is that it connects them to content. >> the network providers >> are in the business of taking a fraction of the business that the >> content providers are in ... > > If you look at any of the research on networks, most researchers agree > that the value of the network is in connectivity. There's arguments > about whether the value scales as O( n**2 ) via Metcalfe's Law or > something more like O( n log(n) ) which Briscoe, Odlyzko, and Tilly > claim. But nobody claims that the value of networks is in the bandwidth. > > Last time I looked, network providers weren't charging for > connectivity, they were charging for bandwidth. Google makes tons of > money off of small text ads that use almost no bandwidth but cash in > from free connectivity. Network providers are forced to give away > connectivity because the Internet architecture provides no way for > them to charge for it. Not a particularly good business when you are > forced to give away what is of value and charge for what isn't. > > My blog post this week discusses this more and the connection with > end-to-end (http://cleanslate-internet.blogspot.com). > > jak > From dwing at cisco.com Tue Apr 22 15:45:08 2008 From: dwing at cisco.com (Dan Wing) Date: Tue, 22 Apr 2008 15:45:08 -0700 Subject: [e2e] SCTP popularity In-Reply-To: References: Message-ID: <00e301c8a4ca$847be400$c2f0200a@cisco.com> Lars Eggert wrote: > On 2008-4-19, at 5:07, ext Alex Moura wrote: > > Inspired by the last thread, I'd like to know your opinions > > regarding SCTP > > adoption and general use. It will never happen, or it will follow > > the same > > process - and time - that IPv6 adoption is taking? > > Adoption - sure; as George has pointed out, IPFIX is using > it, and so do several other applications and application-layer protocols. > > General use as in by you and me behind our home NATs - not > currently. > But the Linux firewall/NAT has SCTP and DCCP support (and BSD has at > least SCTP, I think), and if residential NAT vendors that use those > systems allow enabling that functionality, it may happen. I > grant you that that's a big if, because there is little incentive. There is draft-tuexen-sctp-udp-encaps-02.txt which can help get SCTP over the chicken-and-egg problem of SCTP-unware NATs. -d From Jon.Crowcroft at cl.cam.ac.uk Thu Apr 24 04:59:09 2008 From: Jon.Crowcroft at cl.cam.ac.uk (Jon Crowcroft) Date: Thu, 24 Apr 2008 12:59:09 +0100 Subject: [e2e] experimenting on customers In-Reply-To: <480F296F.7020807@reed.com> References: <48089E95.1090103@reed.com> <480908CC.5050400@reed.com> <480C9C37.40509@reed.com> <480EB31C.7060403@dcrocker.net> <480F296F.7020807@reed.com> Message-ID: it is crucial to realize that the regulation of experiments of any kind is a post hoc rationalisation rather than any kidn of actual model of what ought to be the outcome almost all new succesfully deployed protocols in the last 15-20 years have been ahead of any curve to do with IETF processes, ISPs planning, provisioning, legal system comprehension... end users download and run stuff (even if they dont compromse their OS, they in the millions download facebook apps daily that compromise their privacy and potentially break laws in some parts of the world they upgrde or dont upgrade their end systems and their home xDSL/wifi routers' firmware every one of these may be a controlled experiment when conducted in isolation, and with full support of the software developer, but in combination they are clear blue sky... we don't know the emergent properties of these things until people notice them (in nanog, in court, in government, or, occasionalyl, by doing measurement experiments... frankly, even within a single node, i remember roger needham explaining over 10 years ago that it had become impossible for microsoft to run regression testing across all combinations of devices and drivers and OS versions because the numbers had just Got Too Big already (2-3 new devices per day etc etc) so now do that with networked interactions... ergo: all experiments by net customers are experiments on net customers... of course, the one thing we can't do with the one true internet (since it is now holy critical infrastructure) is proper destrctive testing (we can't even figure out LD50:) In missive <480F296F.7020807 at reed.com>, "David P. Reed" typed: >>Dave - as I mentioned earlier, there is a huge difference between >>experimenting on customers and letting customers experiment. >> >>Your post equates the two. I suggest that the distinction is crucial. >>And that is the point of the end-to-end argument, at least 80% of it, >>when applied to modularity between carriers and users, or the modularity >>between systems vendors and users, or the modularity between companies >>that would hope to support innovative research and researchers. >> >> >> >>Dave Crocker wrote: >>> >>> >>> Jon Crowcroft wrote: >>>> I dont understand all this - most software in the last 30 >>>> years is an experiment on customers - the internet as a whole >>>> is an experiment >>> ... >>>> so if we are honest, we'd admit this and say >>>> what we need is a pharma model of informed consent >>>> yeah, even discounts >>> >>> >>> In looking over the many postings in this thread, the word >>> "experiment" provides the most leverage both for insight and for >>> confusion. >>> >>> Experiments come in very different flavors, notably with very >>> different risk. >>> When talking about something on the scale of the current, public >>> Internet, >>> or American democracy or global jet travel, the term "experiment" >>> reminds us >>> that we do not fully understand impact. But the term also denotes a >>> risk of >>> failure which cannot reasonably apply for these grander uses. (After >>> a few >>> hundred years, if a civilization dies off, is it a "failure", even >>> though we >>> label it an experiment?) In other words, we use the word "experiment" >>> here in >>> a non-technical way, connoting the unknown, rather the denoting >>> controlled >>> manipulation, diligent study and incremental refinement. >>> >>> So, some of the complaints about being unable to experiment on the open >>> Internet simply do not make sense, any more than "testing" a radically >>> new concrete -- with no use experience -- on a freeway bridge would >>> make sense. Risk is obviously too high; in fact, failure early in the >>> lifecycle of a new >>> technology is typically guaranteed. Would you drive over that sucker? >>> Or under it? >>> >>> So if someone is going to express concerns about barriers to adoption, >>> such as >>> a lack of flexibility by providers or product companies, they need to >>> accompany it will a compelling adoption case that shows sufficiently >>> low risk >>> and sufficiently high benefit. Typically, that needs to come from real >>> experimentation, meaning early-stage development, real testing, and pilot >>> deployment. (Quite nicely this has the not-so-minor side benefit of >>> grooming an increasingly significant constituency that wants the >>> technology adopted.) >>> >>> Businesses do not deploy real experiments in their products and services. >>> Costs and risks are far both too high. What they deploy are features >>> that provide relatively assured benefits. >>> >>> As for "blocking" experiments by others, think again of the bridge. >>> Collateral damage requires that public infrastructure services been >>> particularly conservative in permitting change. >>> >>> In the early 90s, when new routing protocols were being defined and >>> debated, it was also noted that there was no 'laboratory' large enough >>> to test the protocols to scale, prior to deployment in the open >>> Internet. One thought was to couple smaller test networks, via >>> tunnels across the Internet. I suppose Internet II counts as a modern >>> alternative. In other words, real experiments needs real laboratories. >>> >>> The other challenge for this particular thread is that the term >>> end-to-end is >>> treated as a rigid absolute, but never has actually been that. It is >>> a term >>> of relativity defined by two boundary points. The the modern Internet >>> has added more complexity between the points, as others have noted. >>> Rather than a simplistic host-net dichotomy we have layers of >>> intermediary nets, and often layers of application hosts. (Thank you, >>> akamai...) We also have layers outside of what we typically call >>> end-points, such as services that treat email as underlying >>> infrastructure, rather than "the" application. >>> >>> And we have layers of trust (tussles). >>> >>> And, and, and... >>> >>> So when claiming to need end-to-end, the question is which of many >>> possible ends? >>> >>> And, for what purpose, or one might say... to what end? >>> >>> d/ >>> >>> >>> >>> >>> >>> >>> >>> >>> >>> cheers jon From dpreed at reed.com Thu Apr 24 07:25:08 2008 From: dpreed at reed.com (David P. Reed) Date: Thu, 24 Apr 2008 10:25:08 -0400 Subject: [e2e] experimenting on customers In-Reply-To: References: <48089E95.1090103@reed.com> <480908CC.5050400@reed.com> <480C9C37.40509@reed.com> <480EB31C.7060403@dcrocker.net> <480F296F.7020807@reed.com> Message-ID: <48109844.2040209@reed.com> So what's the point you're making Jon? Users' experiments impact other users? What does that have to do with vendors experimenting on their users? The cases are different qualitatively and in kind, and the risks and liabilities are different, as in any multiparty system of constraints and desires. Unless of course, you think that there is no difference at all between the King and his subjects, the President and commander-in-chief and a homeless person. Experiments by the powerful upon the weak/dependent are quite different from experiments with limited impact and scale taking place in a vast ocean of relatively unaffected people. There is no binary "good vs. evil" logic here. There is no black and white. But the difference is plain unless you abstract away every aspect of reality other than the term "experiment". Jon Crowcroft wrote: > it is crucial to realize that the regulation of experiments of any kind is a > post hoc rationalisation rather than any kidn of actual model of what > ought to be the outcome > > almost all new succesfully deployed > protocols in the last 15-20 years have been ahead of any curve > to do with IETF processes, ISPs planning, provisioning, > legal system comprehension... > > end users download and run stuff > (even if they dont compromse their OS, they > in the millions download facebook apps daily that compromise their privacy and > potentially break laws in some parts of the world > > they upgrde or dont upgrade their end systems and their home xDSL/wifi routers' > firmware > > every one of these may be a controlled experiment when conducted in isolation, > and with full support of the software developer, but in combination > they are clear > blue sky... > > we don't know the emergent properties of these things until people notice them > (in nanog, in court, in government, or, occasionalyl, by doing measurement > experiments... > > frankly, even within a single node, i remember roger needham explaining over 10 > years ago that it had become impossible for microsoft to run regression testing > across all combinations of devices and drivers and OS versions because the > numbers had just Got Too Big already (2-3 new devices per day etc etc) > so now do that with networked interactions... > > ergo: > all experiments by net customers are experiments on net customers... > > of course, the one thing we can't do with the one true internet (since it is now > holy critical infrastructure) is proper destrctive testing > (we can't even figure out LD50:) > > In missive <480F296F.7020807 at reed.com>, "David P. Reed" typed: > > >>Dave - as I mentioned earlier, there is a huge difference between > >>experimenting on customers and letting customers experiment. > >> > >>Your post equates the two. I suggest that the distinction is crucial. > >>And that is the point of the end-to-end argument, at least 80% of it, > >>when applied to modularity between carriers and users, or the modularity > >>between systems vendors and users, or the modularity between companies > >>that would hope to support innovative research and researchers. > >> > >> > >> > >>Dave Crocker wrote: > >>> > >>> > >>> Jon Crowcroft wrote: > >>>> I dont understand all this - most software in the last 30 > >>>> years is an experiment on customers - the internet as a whole > >>>> is an experiment > >>> ... > >>>> so if we are honest, we'd admit this and say > >>>> what we need is a pharma model of informed consent > >>>> yeah, even discounts > >>> > >>> > >>> In looking over the many postings in this thread, the word > >>> "experiment" provides the most leverage both for insight and for > >>> confusion. > >>> > >>> Experiments come in very different flavors, notably with very > >>> different risk. > >>> When talking about something on the scale of the current, public > >>> Internet, > >>> or American democracy or global jet travel, the term "experiment" > >>> reminds us > >>> that we do not fully understand impact. But the term also denotes a > >>> risk of > >>> failure which cannot reasonably apply for these grander uses. (After > >>> a few > >>> hundred years, if a civilization dies off, is it a "failure", even > >>> though we > >>> label it an experiment?) In other words, we use the word "experiment" > >>> here in > >>> a non-technical way, connoting the unknown, rather the denoting > >>> controlled > >>> manipulation, diligent study and incremental refinement. > >>> > >>> So, some of the complaints about being unable to experiment on the open > >>> Internet simply do not make sense, any more than "testing" a radically > >>> new concrete -- with no use experience -- on a freeway bridge would > >>> make sense. Risk is obviously too high; in fact, failure early in the > >>> lifecycle of a new > >>> technology is typically guaranteed. Would you drive over that sucker? > >>> Or under it? > >>> > >>> So if someone is going to express concerns about barriers to adoption, > >>> such as > >>> a lack of flexibility by providers or product companies, they need to > >>> accompany it will a compelling adoption case that shows sufficiently > >>> low risk > >>> and sufficiently high benefit. Typically, that needs to come from real > >>> experimentation, meaning early-stage development, real testing, and pilot > >>> deployment. (Quite nicely this has the not-so-minor side benefit of > >>> grooming an increasingly significant constituency that wants the > >>> technology adopted.) > >>> > >>> Businesses do not deploy real experiments in their products and services. > >>> Costs and risks are far both too high. What they deploy are features > >>> that provide relatively assured benefits. > >>> > >>> As for "blocking" experiments by others, think again of the bridge. > >>> Collateral damage requires that public infrastructure services been > >>> particularly conservative in permitting change. > >>> > >>> In the early 90s, when new routing protocols were being defined and > >>> debated, it was also noted that there was no 'laboratory' large enough > >>> to test the protocols to scale, prior to deployment in the open > >>> Internet. One thought was to couple smaller test networks, via > >>> tunnels across the Internet. I suppose Internet II counts as a modern > >>> alternative. In other words, real experiments needs real laboratories. > >>> > >>> The other challenge for this particular thread is that the term > >>> end-to-end is > >>> treated as a rigid absolute, but never has actually been that. It is > >>> a term > >>> of relativity defined by two boundary points. The the modern Internet > >>> has added more complexity between the points, as others have noted. > >>> Rather than a simplistic host-net dichotomy we have layers of > >>> intermediary nets, and often layers of application hosts. (Thank you, > >>> akamai...) We also have layers outside of what we typically call > >>> end-points, such as services that treat email as underlying > >>> infrastructure, rather than "the" application. > >>> > >>> And we have layers of trust (tussles). > >>> > >>> And, and, and... > >>> > >>> So when claiming to need end-to-end, the question is which of many > >>> possible ends? > >>> > >>> And, for what purpose, or one might say... to what end? > >>> > >>> d/ > >>> > >>> > >>> > >>> > >>> > >>> > >>> > >>> > >>> > >>> > > cheers > > jon > > > From dhc2 at dcrocker.net Thu Apr 24 10:16:34 2008 From: dhc2 at dcrocker.net (Dave Crocker) Date: Thu, 24 Apr 2008 10:16:34 -0700 Subject: [e2e] experimenting on customers In-Reply-To: References: <48089E95.1090103@reed.com> <480908CC.5050400@reed.com> <480C9C37.40509@reed.com> <480EB31C.7060403@dcrocker.net> <480F296F.7020807@reed.com> Message-ID: <4810C072.5060107@dcrocker.net> Jon Crowcroft wrote: > ergo: > all experiments by net customers are experiments on net customers... > > of course, the one thing we can't do with the one true internet (since it is now > holy critical infrastructure) is proper destrctive testing And this points to the activity that seems not to have been listed: experimenting *on* infrastructure. Which is exactly the case that is blocked for users and mostly for providers. As it should be. A provider might be "allowed" to experiment on their own infrastructure, but any competent operator will take a particularly conservative approach, given the downside of a failure. But the idea that a provider should let a customer conduct an experiment that would affect the infrastructure doesn't sound reasonable. Whether a particular product allows users to experiment depends on the product. One intended for the consumer market typically won't. To the extent that that makes the product developer a gatekeeper for the acceptance -- not the experimentation -- of a new application, well, welcome to the real world. So we need to distinguish among who the experimenter is, what they are experimenting on, what they are experimenting with, and what kinds of collateral damage are possible. Does a provider allow a customer to experiment with a new application? Typically yes, though indeed, some providers are Draconian. Does an o/s platform company allow a customer experiment with a new application? It depends on the platform. Typically one that is tailored for the mass market is constrained in order to keep things simple. (Whether it succeeds in that goal is a different discussion.) And to go back to Jon's observation that none of these 'experiments' is fully controlled and measured... sure. But there are activities that are conducted with an expectation of problems and with a focus on finding them. Interoperability events are examples. If the impact of failure is high enough, then these events are always conducted in isolated networks. Providers do not get in the way of interoperability events for applications over the public Internet. They *do* get in the way of experimentation with infrastructure services, such as replacing IPv4. And they should. All of which seems to lead back to a question about the complaints being aired. Exactly what activities are being prevented? How is their being prevented the result of inappropriate actions? d/ -- Dave Crocker Brandenburg InternetWorking bbiw.net From Jon.Crowcroft at cl.cam.ac.uk Fri Apr 25 00:33:15 2008 From: Jon.Crowcroft at cl.cam.ac.uk (Jon Crowcroft) Date: Fri, 25 Apr 2008 08:33:15 +0100 Subject: [e2e] experimenting on customers In-Reply-To: <48109844.2040209@reed.com> References: <48089E95.1090103@reed.com> <480908CC.5050400@reed.com> <480C9C37.40509@reed.com> <480EB31C.7060403@dcrocker.net> <480F296F.7020807@reed.com> <48109844.2040209@reed.com> Message-ID: the point is that very very small people can do very very big experiments - there was some controversay about this, for example, in NSI Last year when the bittyrant people reveleaved that they had released their variant of the torrent tool with a modified incentive algorthm to see what would happen with a lot of users - as with all good psycho-science (and some anthropolgy:), the users can't know you are doing the experiment, coz that might interfere with the validity (sounds like asimov's psychohistory;)..... but of course that has interesting ethical impact... but thats not my main point, which is: something as small as few lines change in a p2p program which is then run by 1000s or millions of users, has a MASSIVE potential (and actual) impact on the traffic pattern, which has a massive impact on the ISPs (infrastructure) which has a massive impact on the other users. so just because you cannot alter an IP header or a TCP option saying a) the middleboxes get in the way, and b) the vendors wont put it in the OS stack for you anyhow, does NOT mean you cannot do BIG Network Science one bit. not at all. oh no In missive <48109844.2040209 at reed.com>, "David P. Reed" typed: >>So what's the point you're making Jon? Users' experiments impact other >>users? What does that have to do with vendors experimenting on their >>users? The cases are different qualitatively and in kind, and the risks >>and liabilities are different, as in any multiparty system of >>constraints and desires. >> >>Unless of course, you think that there is no difference at all between >>the King and his subjects, the President and commander-in-chief and a >>homeless person. >> >>Experiments by the powerful upon the weak/dependent are quite different >>from experiments with limited impact and scale taking place in a vast >>ocean of relatively unaffected people. >> >>There is no binary "good vs. evil" logic here. There is no black and >>white. But the difference is plain unless you abstract away every >>aspect of reality other than the term "experiment". >> >>Jon Crowcroft wrote: >>> it is crucial to realize that the regulation of experiments of any kind is a >>> post hoc rationalisation rather than any kidn of actual model of what >>> ought to be the outcome >>> >>> almost all new succesfully deployed >>> protocols in the last 15-20 years have been ahead of any curve >>> to do with IETF processes, ISPs planning, provisioning, >>> legal system comprehension... >>> >>> end users download and run stuff >>> (even if they dont compromse their OS, they >>> in the millions download facebook apps daily that compromise their privacy and >>> potentially break laws in some parts of the world >>> >>> they upgrde or dont upgrade their end systems and their home xDSL/wifi routers' >>> firmware >>> >>> every one of these may be a controlled experiment when conducted in isolation, >>> and with full support of the software developer, but in combination >>> they are clear >>> blue sky... >>> >>> we don't know the emergent properties of these things until people notice them >>> (in nanog, in court, in government, or, occasionalyl, by doing measurement >>> experiments... >>> >>> frankly, even within a single node, i remember roger needham explaining over 10 >>> years ago that it had become impossible for microsoft to run regression testing >>> across all combinations of devices and drivers and OS versions because the >>> numbers had just Got Too Big already (2-3 new devices per day etc etc) >>> so now do that with networked interactions... >>> >>> ergo: >>> all experiments by net customers are experiments on net customers... >>> >>> of course, the one thing we can't do with the one true internet (since it is now >>> holy critical infrastructure) is proper destrctive testing >>> (we can't even figure out LD50:) >>> >>> In missive <480F296F.7020807 at reed.com>, "David P. Reed" typed: >>> >>> >>Dave - as I mentioned earlier, there is a huge difference between >>> >>experimenting on customers and letting customers experiment. >>> >> >>> >>Your post equates the two. I suggest that the distinction is crucial. >>> >>And that is the point of the end-to-end argument, at least 80% of it, >>> >>when applied to modularity between carriers and users, or the modularity >>> >>between systems vendors and users, or the modularity between companies >>> >>that would hope to support innovative research and researchers. >>> >> >>> >> >>> >> >>> >>Dave Crocker wrote: >>> >>> >>> >>> >>> >>> Jon Crowcroft wrote: >>> >>>> I dont understand all this - most software in the last 30 >>> >>>> years is an experiment on customers - the internet as a whole >>> >>>> is an experiment >>> >>> ... >>> >>>> so if we are honest, we'd admit this and say >>> >>>> what we need is a pharma model of informed consent >>> >>>> yeah, even discounts >>> >>> >>> >>> >>> >>> In looking over the many postings in this thread, the word >>> >>> "experiment" provides the most leverage both for insight and for >>> >>> confusion. >>> >>> >>> >>> Experiments come in very different flavors, notably with very >>> >>> different risk. >>> >>> When talking about something on the scale of the current, public >>> >>> Internet, >>> >>> or American democracy or global jet travel, the term "experiment" >>> >>> reminds us >>> >>> that we do not fully understand impact. But the term also denotes a >>> >>> risk of >>> >>> failure which cannot reasonably apply for these grander uses. (After >>> >>> a few >>> >>> hundred years, if a civilization dies off, is it a "failure", even >>> >>> though we >>> >>> label it an experiment?) In other words, we use the word "experiment" >>> >>> here in >>> >>> a non-technical way, connoting the unknown, rather the denoting >>> >>> controlled >>> >>> manipulation, diligent study and incremental refinement. >>> >>> >>> >>> So, some of the complaints about being unable to experiment on the open >>> >>> Internet simply do not make sense, any more than "testing" a radically >>> >>> new concrete -- with no use experience -- on a freeway bridge would >>> >>> make sense. Risk is obviously too high; in fact, failure early in the >>> >>> lifecycle of a new >>> >>> technology is typically guaranteed. Would you drive over that sucker? >>> >>> Or under it? >>> >>> >>> >>> So if someone is going to express concerns about barriers to adoption, >>> >>> such as >>> >>> a lack of flexibility by providers or product companies, they need to >>> >>> accompany it will a compelling adoption case that shows sufficiently >>> >>> low risk >>> >>> and sufficiently high benefit. Typically, that needs to come from real >>> >>> experimentation, meaning early-stage development, real testing, and pilot >>> >>> deployment. (Quite nicely this has the not-so-minor side benefit of >>> >>> grooming an increasingly significant constituency that wants the >>> >>> technology adopted.) >>> >>> >>> >>> Businesses do not deploy real experiments in their products and services. >>> >>> Costs and risks are far both too high. What they deploy are features >>> >>> that provide relatively assured benefits. >>> >>> >>> >>> As for "blocking" experiments by others, think again of the bridge. >>> >>> Collateral damage requires that public infrastructure services been >>> >>> particularly conservative in permitting change. >>> >>> >>> >>> In the early 90s, when new routing protocols were being defined and >>> >>> debated, it was also noted that there was no 'laboratory' large enough >>> >>> to test the protocols to scale, prior to deployment in the open >>> >>> Internet. One thought was to couple smaller test networks, via >>> >>> tunnels across the Internet. I suppose Internet II counts as a modern >>> >>> alternative. In other words, real experiments needs real laboratories. >>> >>> >>> >>> The other challenge for this particular thread is that the term >>> >>> end-to-end is >>> >>> treated as a rigid absolute, but never has actually been that. It is >>> >>> a term >>> >>> of relativity defined by two boundary points. The the modern Internet >>> >>> has added more complexity between the points, as others have noted. >>> >>> Rather than a simplistic host-net dichotomy we have layers of >>> >>> intermediary nets, and often layers of application hosts. (Thank you, >>> >>> akamai...) We also have layers outside of what we typically call >>> >>> end-points, such as services that treat email as underlying >>> >>> infrastructure, rather than "the" application. >>> >>> >>> >>> And we have layers of trust (tussles). >>> >>> >>> >>> And, and, and... >>> >>> >>> >>> So when claiming to need end-to-end, the question is which of many >>> >>> possible ends? >>> >>> >>> >>> And, for what purpose, or one might say... to what end? >>> >>> >>> >>> d/ >>> >>> >>> >>> >>> >>> >>> >>> >>> >>> >>> >>> >>> >>> >>> >>> >>> >>> >>> >>> >>> >>> cheers >>> >>> jon >>> >>> >>> cheers jon From dpreed at reed.com Fri Apr 25 14:21:03 2008 From: dpreed at reed.com (David P. Reed) Date: Fri, 25 Apr 2008 17:21:03 -0400 Subject: [e2e] experimenting on customers In-Reply-To: References: <48089E95.1090103@reed.com> <480908CC.5050400@reed.com> <480C9C37.40509@reed.com> <480EB31C.7060403@dcrocker.net> <480F296F.7020807@reed.com> <48109844.2040209@reed.com> Message-ID: <48124B3F.3050702@reed.com> There is no doubt that individuals and small groups can have big effects. Ken Thompson and a couple others created Unix that disentangled the idea of an operating system from particular hardware in a few months of inspired hacking, Dan Bricklin and Bob Frankston launched the business use of personal computing with a few months coding in Bob's attic, and 30-40 grad students across the Northern Hemisphere launched the Internet as an overlay network that changed the world. Such small experiments, which scale by explicitly exploiting, without need to "ask permission", others' technology, need to be encouraged, not discouraged. Encouraged even though they "gore the oxen" of vendors of the exploited technologies, and threaten to exceed the bounds of control desired by control-freak governments, fears promoted by those who dislike new ideas and unpredictability. We NEED fast, cheap, and out-of-control inventions. Those inventions are the only source of results that can deal with the equally dangerous exponential growth of populations, pollution, climate risks, etc. With all the attendant risks and need for mindfulness on the part of the inventors themselves that attend upon such powerful agencies of change. Jon Crowcroft wrote: > the point is that very very small people can do very very big experiments - there > was some controversay about this, for example, in NSI Last year when the > bittyrant people reveleaved that they had released their variant of the torrent > tool with a modified incentive algorthm to see what would happen with a lot of > users - as with all good psycho-science (and some anthropolgy:), the users > can't know you are doing the experiment, coz that might interfere with the > validity (sounds like asimov's psychohistory;)..... > but of course that has interesting ethical impact... > > but thats not my main point, which is: > > something as small as few lines change in a p2p program > which is then run by 1000s or millions of users, > has a MASSIVE potential (and actual) impact on the traffic pattern, > which has a massive impact on the ISPs (infrastructure) > which has a massive impact on the other users. > > so just because you cannot alter an IP header or a TCP option saying > a) the middleboxes get in the way, > and > b) the vendors wont put it in the OS stack for you anyhow, > does NOT mean you cannot do BIG Network Science > one bit. not at all. > > oh no > > > In missive <48109844.2040209 at reed.com>, "David P. Reed" typed: > > >>So what's the point you're making Jon? Users' experiments impact other > >>users? What does that have to do with vendors experimenting on their > >>users? The cases are different qualitatively and in kind, and the risks > >>and liabilities are different, as in any multiparty system of > >>constraints and desires. > >> > >>Unless of course, you think that there is no difference at all between > >>the King and his subjects, the President and commander-in-chief and a > >>homeless person. > >> > >>Experiments by the powerful upon the weak/dependent are quite different > >>from experiments with limited impact and scale taking place in a vast > >>ocean of relatively unaffected people. > >> > >>There is no binary "good vs. evil" logic here. There is no black and > >>white. But the difference is plain unless you abstract away every > >>aspect of reality other than the term "experiment". > >> > >>Jon Crowcroft wrote: > >>> it is crucial to realize that the regulation of experiments of any kind is a > >>> post hoc rationalisation rather than any kidn of actual model of what > >>> ought to be the outcome > >>> > >>> almost all new succesfully deployed > >>> protocols in the last 15-20 years have been ahead of any curve > >>> to do with IETF processes, ISPs planning, provisioning, > >>> legal system comprehension... > >>> > >>> end users download and run stuff > >>> (even if they dont compromse their OS, they > >>> in the millions download facebook apps daily that compromise their privacy and > >>> potentially break laws in some parts of the world > >>> > >>> they upgrde or dont upgrade their end systems and their home xDSL/wifi routers' > >>> firmware > >>> > >>> every one of these may be a controlled experiment when conducted in isolation, > >>> and with full support of the software developer, but in combination > >>> they are clear > >>> blue sky... > >>> > >>> we don't know the emergent properties of these things until people notice them > >>> (in nanog, in court, in government, or, occasionalyl, by doing measurement > >>> experiments... > >>> > >>> frankly, even within a single node, i remember roger needham explaining over 10 > >>> years ago that it had become impossible for microsoft to run regression testing > >>> across all combinations of devices and drivers and OS versions because the > >>> numbers had just Got Too Big already (2-3 new devices per day etc etc) > >>> so now do that with networked interactions... > >>> > >>> ergo: > >>> all experiments by net customers are experiments on net customers... > >>> > >>> of course, the one thing we can't do with the one true internet (since it is now > >>> holy critical infrastructure) is proper destrctive testing > >>> (we can't even figure out LD50:) > >>> > >>> In missive <480F296F.7020807 at reed.com>, "David P. Reed" typed: > >>> > >>> >>Dave - as I mentioned earlier, there is a huge difference between > >>> >>experimenting on customers and letting customers experiment. > >>> >> > >>> >>Your post equates the two. I suggest that the distinction is crucial. > >>> >>And that is the point of the end-to-end argument, at least 80% of it, > >>> >>when applied to modularity between carriers and users, or the modularity > >>> >>between systems vendors and users, or the modularity between companies > >>> >>that would hope to support innovative research and researchers. > >>> >> > >>> >> > >>> >> > >>> >>Dave Crocker wrote: > >>> >>> > >>> >>> > >>> >>> Jon Crowcroft wrote: > >>> >>>> I dont understand all this - most software in the last 30 > >>> >>>> years is an experiment on customers - the internet as a whole > >>> >>>> is an experiment > >>> >>> ... > >>> >>>> so if we are honest, we'd admit this and say > >>> >>>> what we need is a pharma model of informed consent > >>> >>>> yeah, even discounts > >>> >>> > >>> >>> > >>> >>> In looking over the many postings in this thread, the word > >>> >>> "experiment" provides the most leverage both for insight and for > >>> >>> confusion. > >>> >>> > >>> >>> Experiments come in very different flavors, notably with very > >>> >>> different risk. > >>> >>> When talking about something on the scale of the current, public > >>> >>> Internet, > >>> >>> or American democracy or global jet travel, the term "experiment" > >>> >>> reminds us > >>> >>> that we do not fully understand impact. But the term also denotes a > >>> >>> risk of > >>> >>> failure which cannot reasonably apply for these grander uses. (After > >>> >>> a few > >>> >>> hundred years, if a civilization dies off, is it a "failure", even > >>> >>> though we > >>> >>> label it an experiment?) In other words, we use the word "experiment" > >>> >>> here in > >>> >>> a non-technical way, connoting the unknown, rather the denoting > >>> >>> controlled > >>> >>> manipulation, diligent study and incremental refinement. > >>> >>> > >>> >>> So, some of the complaints about being unable to experiment on the open > >>> >>> Internet simply do not make sense, any more than "testing" a radically > >>> >>> new concrete -- with no use experience -- on a freeway bridge would > >>> >>> make sense. Risk is obviously too high; in fact, failure early in the > >>> >>> lifecycle of a new > >>> >>> technology is typically guaranteed. Would you drive over that sucker? > >>> >>> Or under it? > >>> >>> > >>> >>> So if someone is going to express concerns about barriers to adoption, > >>> >>> such as > >>> >>> a lack of flexibility by providers or product companies, they need to > >>> >>> accompany it will a compelling adoption case that shows sufficiently > >>> >>> low risk > >>> >>> and sufficiently high benefit. Typically, that needs to come from real > >>> >>> experimentation, meaning early-stage development, real testing, and pilot > >>> >>> deployment. (Quite nicely this has the not-so-minor side benefit of > >>> >>> grooming an increasingly significant constituency that wants the > >>> >>> technology adopted.) > >>> >>> > >>> >>> Businesses do not deploy real experiments in their products and services. > >>> >>> Costs and risks are far both too high. What they deploy are features > >>> >>> that provide relatively assured benefits. > >>> >>> > >>> >>> As for "blocking" experiments by others, think again of the bridge. > >>> >>> Collateral damage requires that public infrastructure services been > >>> >>> particularly conservative in permitting change. > >>> >>> > >>> >>> In the early 90s, when new routing protocols were being defined and > >>> >>> debated, it was also noted that there was no 'laboratory' large enough > >>> >>> to test the protocols to scale, prior to deployment in the open > >>> >>> Internet. One thought was to couple smaller test networks, via > >>> >>> tunnels across the Internet. I suppose Internet II counts as a modern > >>> >>> alternative. In other words, real experiments needs real laboratories. > >>> >>> > >>> >>> The other challenge for this particular thread is that the term > >>> >>> end-to-end is > >>> >>> treated as a rigid absolute, but never has actually been that. It is > >>> >>> a term > >>> >>> of relativity defined by two boundary points. The the modern Internet > >>> >>> has added more complexity between the points, as others have noted. > >>> >>> Rather than a simplistic host-net dichotomy we have layers of > >>> >>> intermediary nets, and often layers of application hosts. (Thank you, > >>> >>> akamai...) We also have layers outside of what we typically call > >>> >>> end-points, such as services that treat email as underlying > >>> >>> infrastructure, rather than "the" application. > >>> >>> > >>> >>> And we have layers of trust (tussles). > >>> >>> > >>> >>> And, and, and... > >>> >>> > >>> >>> So when claiming to need end-to-end, the question is which of many > >>> >>> possible ends? > >>> >>> > >>> >>> And, for what purpose, or one might say... to what end? > >>> >>> > >>> >>> d/ > >>> >>> > >>> >>> > >>> >>> > >>> >>> > >>> >>> > >>> >>> > >>> >>> > >>> >>> > >>> >>> > >>> >>> > >>> > >>> cheers > >>> > >>> jon > >>> > >>> > >>> > > cheers > > jon > > > From Jon.Crowcroft at cl.cam.ac.uk Sat Apr 26 03:18:29 2008 From: Jon.Crowcroft at cl.cam.ac.uk (Jon Crowcroft) Date: Sat, 26 Apr 2008 11:18:29 +0100 Subject: [e2e] dynamic ISP selection Message-ID: so in the rest of the world where we have competition in the last mile (i.e. not the US:), it is surprisingly tricky to change ISP - even though some massive fraction of xDSL is unbundled - [and the regulators looked recently and discovered that the widespread competition between cable modem and (surpirsingly) broadband wireless access (e.g. umts or high speed packet, or even wimax a bit) means that you get techno-disversity as well as economic diversity in the UK for example, there's something like 35M phone lines that bt owns copper up to the exchange building, but then they get switced onto a humungous ATM net ('colossus'), and can pop out at layer to below IP in any virtual broadband provider's POP (asde: this business is almost like the way that some cell phone service providers work who don't actually own spectrum, but make a business out of leasing spectrum off of others- i think Virgin does this with (maybe) t-mobile's )...their businesses are basically being creative with contract models....i.e. combining other services and content... so anyhow, in the uk (sorry to be so blighty-centric) you can change your utilities pretty much once a month (I just changed gas&electricity twice in the last two months to get better deals) as the same virtualisation (unbundling) of last mile is done, and the core provider (actually 2 layers- grid and generation) are seperated cleanly... so given i can go on the web and get my gas, electricity, water, sewage changed , why is it not possible to get a SPOT price for broadband internet? i cannot see any technical barrier to this whatsoever:) j.