William Herrin allegedly wrote on 11/21/2009 8:41 AM: > On Wed, Nov 18, 2009 at 5:00 AM, Scott Brim <scott.b...@gmail.com> wrote: >> Excerpts from William Herrin on Mon, Nov 16, 2009 09:24:36PM -0500: >>> 2. Host modification. The inbound NAT adds an IP extension with the >>> original destination address. The host echos this in the return >>> packet, providing the outbound NAT with the info he needs to set the >>> correct external source address. >> Another possibility is to decouple establishment of higher layer state >> from the IP locators the packets come tagged with. >> >> For example, if they are using TCP and the correct tokens are provided >> with SYN etc., then the result of having packets sourced from >> "unexpected" locators is that endpoints learn new locators that could >> be used for the session -- not that the session can't be initiated. > > Hi Scott, > > That would be a host modification.
Yes. Hosts are being modified anyway, regardless of what RRG does or does not do. As long as routing scaling does not _depend_ on host modification, it can _assume_ endpoint evolution in architectural planning. > Speaking of which, has anyone in > this WG considered setting aside the great debate about architectures > long enough to try to compose some consensus recommendations targetted > at the Multipath TCP working group stating what exactly the routing > folks would like to see them achieve? That's what I was suggesting should be included in RRG's "recommendation" to the IETF :-) -- a set of engineering boundaries the IETF has to decide on. There has already been lots of ad hoc discussion. In the end it's essentially decoupled. The majority of Internet endpoints will be mobile, and they cannot assume any particular routing and addressing regime in whatever networks they connect to, so they will take care of themselves. > If we'd like them to carry a > decoupled session ID in every packet and deal with it successfully > when the IP address for a particular flow changes unexpectedly, we > might want to mention that before they finish. Do you think routing and addressing requires a session ID in every packet? If not, let the upper layers find their own solutions -- e2e argument and all that. Scott _______________________________________________ rrg mailing list rrg@irtf.org http://www.irtf.org/mailman/listinfo/rrg
