[e2e] end2end-interest Digest, Vol 39, Issue 32

Peyman Faratin peyman at MIT.EDU
Tue May 22 11:23:11 PDT 2007

No, it does not make me feel superior. In fact it has a negative  
effect to see that after 35 years networking community has failed to  
produce thoughtful insights in a fundamental problem as economics and  
regulation of networks. The arguments offered are often quasi- 
economic, ad-hoc, private and inconsistent in both terminology and  
historical facts.  Not scientific, and more akin to Scientology.  
Economics and regulation of institutions (not just markets) has  
preoccupied society for centuries and concepts of natural monopoly,  
externalities, elasticities, return to scale, incentives,  
discrimination, utility, costs, marginalism, institutions, risk,  
market structure, market power, etc....have come to represent  
phenomena, and captured as primitives of models, that have a _shared_  
meaning that permits some semblance of scientific method to be  
applied. Clearly, economics is not complete, has methodological  
problems, and has a long way to go, but as academics it is our  
responsibility to bring clarity to problems through the scientific  
method. Your posting was not only ambiguous in its usage of the  
concepts, but was also incomplete and incorrect in its interpretation  
of historical facts. As I mentioned there was asymmetric regulation  
of PSTN-IP overlay settlements that helped IP in the early days  
before infrastructure-based competition could begin and the new sunk  
cost economics could begin to play out. In fact even this is an  
incomplete picture. Usage insensitive pricing, interconnection  
standards (at layer 2 and 3), installed base of elastic applications  
(email and http),  economies of scale of transit contracts,  
competitive backbones, emerging competition in access technology  
markets are just few other first-order reasons why IP scaled, in  
addition to the relative marginal costs induced by asymmetric  
regulation. Some even argue that lack of security was a first-order  
contributor to scaling. My post was intended to bring clarity to the  
discussion and provide references to those interested, and not be an  
attack.  I simply do not share many of your economic conclusions.

Ontology is the first place disciplines meet. Economic colleagues  
tell me that anti-trust lawyers have now after two or so decades come  
to finally understand the _basic_ concept of price discrimination.  
The truth is that the science of network economics (or for that  
matter any substantive discipline) is not unlike the economics of the  
infrastructure itself - it has high fixed costs of learning the  
science, a cost which is often not recoverable (i.e it is sunk) given  
CS departments do not incentivize truly multi-disciplinary research.  
So people substitute for low cost alternatives.

It was never my intention to be flaming people on this list. It is a  
reflection on the state of our science, or lack thereof.



p.s. again, some useful text

[1] H. E. Nuechterlein and P.J. Weiser (2005) Digital Crossroads:  
American Telecommunications Policy in the Internet Age, MIT Press,  
Cambridge, MA, US, 2005

[2] Handbook of Telecommunications Economics, Technology Evolution  
and the Internet, Vol.2, S.K. Majumdar, I Vogelsang and M. Cave  
(eds), Elsevier, 289—364, 2005.

[3] J. Tirole (1988): Industrial Organization, MIT Press, Cambridge,  

[4] J.J. Laffont and J. Tirole (2001): Competition in Telecommunication.

On May 21, 2007, at 5:43 PM, David P. Reed wrote:

> Insulting the Media Lab as a playground is rather unnecessary, and  
> since I claimed no credibility from where I work, it really sounds  
> rather stupid as a rhetorical device.  Does it make you feel superior?
> I got my terminology regarding network externalities and increasing  
> returns from discussions and writings with economists and business  
> school professors.  It's very possible that I'm wrong in my usage -  
> and I'm happy to be corrected.
> However, I didn't claim that network externalities were the *same*  
> as increasing returns to scale.   My fault for a non-parallel  
> construction in that sentence, which might suggest that I thought  
> that they were the same thing.   They are different, and *both*  
> apply to my argument.
> I was using the economist term "network externalities" not the  
> regulatory law term.   Sorry to be confusing if I was.   If you are  
> instead taking the opportunity for gratuitous insult, my skin is  
> thick.
> Peyman Faratin wrote:
>> David
>> I am not sure whether the folks are building computers from sand  
>> in the playground of the media lab, but I can say that you are  
>> "inventing your own science from scratch".
>> Network externality is _not_ the same concept as increasing return  
>> to scale. One is to do with (desirable/undesirable) side-effects  
>> of actions of one agent on another not in the original contract  
>> (externalities) and the other is to do with efficiency gains of  
>> productions of a _single_ firm (where the average cost of  
>> production diminishes with increasing quantities produced). The  
>> marginal cost of checking these facts is insignificant in this day  
>> and age of google and wikipedia.
>> Regulation and economics of networks are non-trivial and require  
>> attention to the details of the arguments that people so freely  
>> misrepresent. Government regulation did have a very significant  
>> impact on the Internet (through differential settlement structures  
>> between interconnecting PSTN and early dialup ISPs). This  
>> government regulation of settlements in fact _helped_ Internet  
>> scaling, not to mention their public investment in the  
>> interchanges and the backbone. MSN was rolled out and could've  
>> tipped to become the dominant standard (as many other inferior  
>> technologies have done so historically - VHS/Betmax, Gasoline/ 
>> steam,....). See [1] and [2] for some more recent text on the  
>> economics and regulation of Internet.
>> Determination of causality in an (un)regulated economy is a very  
>> non-trivial task and, like all sciences, the validity of an  
>> economic (and the accompanying regulatory) hypothesis/proposition  
>> is conditioned on the semantics of the model primitives  
>> (externalities, returns to scale, etc). The devil is in the details.
>> best
>> Peyman
>> [1] H. E. Nuechterlein and P.J. Weiser (2005) /Digital Crossroads:  
>> American Telecommunications Policy in the Internet Age/, MIT  
>> Press, Cambridge, MA, US, 2005
>> [2] Handbook of Telecommunications Economics, Technology Evolution  
>> and the Internet, Vol.2, S.K. Majumdar, I Vogelsang and M. Cave  
>> (eds), Elsevier, 289—364, 2005.
>>> One can, at any time, create a non-interoperable network.    
>>> Corporations do it all the time - they create a boundary at their  
>>> corporate edge and block transit, entry, and exit of packets.
>>> That is not the Internet.   It's a private network based on the  
>>> IP protocol stack.  Things get muddier if there is limited  
>>> interconnection.   Then, the Internet would be best defined as  
>>> the bounded system of endpoints that can each originate packets  
>>> to each other that *will* be delivered with best efforts.   It's  
>>> hard to draw that boundary, but it exists.
>>> Given this view, I don't think government regulations played a  
>>> significant role in the growth of the Internet.   We have had  
>>> lots of private networks, many larger than the Internet.   I  
>>> know, for example, that Microsoft built a large x.25 based  
>>> network in 1985 to provide non-Internet dialup information  
>>> services for Windows.   It was called MSN, and was designed to  
>>> compete with the large private AOL network.
>>> What the Internet had going for it was *scalability* and  
>>> *interoperability*.   Metcalfe's Law and Reed's Law and other  
>>> "laws".   Those created connectivity options that scaled faster  
>>> than private networks could.   Economists calle these "network  
>>> externalites"  or "increasing returns to scale".
>>> Gov't regulation *could* have killed the Internet's  
>>> scalability.   Easy ways would be to make interconnection of  
>>> networks a point of control that was taxed or monitored (e.g. if  
>>> trans-border connectivity were viewed as a point for US Customs  
>>> do inspect or if CALEA were implemented at peering points).
>>> But lacking that, AOL and MSN just could not compete with the  
>>> scalability of the Internet.
>>> That has nothing to do with competition to supply IP software  
>>> stacks in operating systems, or competition among Ethernet  
>>> hardware vendors.
>>> However, increasing returns to scale is not Destiny.   The  
>>> Internet was not destined to become the sole network.   But all  
>>> the members of the Internet (people who can communicate with  
>>> anyone else on it) would be nuts to choose a lesser network  
>>> unless they suffer badly enough to outweigh the collective benefits.
>>> Individualist hermits don't get this, I guess.   If you want to  
>>> be left alone, and don't depend on anyone else, there are no  
>>> returns to scale for you at any scale.   Grow your own bits in  
>>> the backyard, make your own computers from sand, invent your own  
>>> science from scratch.   If the walls are high enough, perhaps you  
>>> can survive without connectivity.

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mailman.postel.org/pipermail/end2end-interest/attachments/20070522/c516415a/attachment-0001.html

More information about the end2end-interest mailing list