On 2009-10-28 05:27, Scott Brim wrote: > Eliot Lear allegedly wrote on 10/27/2009 12:12 PM: >> On 10/26/09 3:36 PM, Scott Brim wrote: >>> - NAT is now architecture. The IETF needs to decide how much NAT it >>> wants the Internet to have in the future. This strongly influences >>> what we do in routing/addressing, because outlying, dwindling cases >>> (either the NAT ones or the non-NAT ones) can be handled specially. >> Do people believe that a scalable bidirectional connection model is >> something that the layer 3 architecture (a) should support and (b) does >> support under the above constraint? > > If I understand correctly, I believe it is fundamentally important. > However, it can be handled through various dynamic lookup services (e.g. > DNS), no?
Well, no. The argument is quite complex, partly because "bidirectional" isn't the general case. I have to point to my own draft: http://tools.ietf.org/html/draft-carpenter-behave-referral-object where we talk about address scope in the presence of middleboxes. I really don't want summarise the argument in a few words, but let me advertise the GROBJ BOF in Hirosahima while I'm here. Brian _______________________________________________ rrg mailing list [email protected] http://www.irtf.org/mailman/listinfo/rrg
