[e2e] Lost Layer?

Matta, Abraham I matta at bu.edu
Tue Jan 14 06:51:36 PST 2014

It's not only "private" addresses within each scope (layer), it's also
that addresses identify processes (and not interfaces as IP addresses do.)
By late binding a process address to a particular interface, one can
better deal with mobility and multihoming.

Each layer also provides transport (flow) services, which can depend on
lower layers (i.e. a recursive model). Each layer basically implements its
own control loop over its scope. Functions like addressing, routing, flow
control, etc. are implemented recursively (http://csr.bu.edu/rina/). In
the current Internet, the IP layer has a huge scope (range of operation)
which makes it very hard to control (from control theory, we know how hard
it is to stabilize a feedback system when the feedback delay is large).


>On 1/13/14 12:57 PM, "Steven Blake" <slblake at petri-meat.com> wrote:
>>On Sat, 2014-01-11 at 11:40 +0000, Jon Crowcroft wrote:
>>> certainly the authors should have read IEN1 which would have led to a
>>> much better identifier space (as revived in the ILNP work by ran
>>> atkinson, 40 years later
>>> http://ilnp.cs.st-andrews.ac.uk/
>>>  cheers
>>>    jon
>>If I understand the slides correctly, they argue that what is missing
>>are network-scope (e.g., AS-scope) addresses, to be used by
>>hosts/routers within a domain, restricting the use (e.g.,
>>routing/forwarding) of Internet-scope addresses to hosts and Internet
>>gateways (e.g., inter-exchange routers).  Whatever the other merits of
>>that approach, it won't solve the multihoming scalability problem.
>>// Steve

More information about the end2end-interest mailing list