Brian, This is a good argument. However, the PI/PA issue may be the elephant in the room in this entire discussion. It is certainly a key issue that the RRG WG debated. Given that, I would like to see words in the recommendation that explicitly acknowledge that routing scaling benefits from locator aggregation while simultaneously end-users' nodal identifiers should be able to not have any ISP dependencies. The specific technology which is used to decouple the historic semantic overloading of the "IP identity problem" is less important for this WG than the fact that decoupling somehow needs to be able to occur via some appropriate technology choice for the local environment. To satisfy everyone, why not just enumerate all of the current possibilities (IRON-RANGER, SHIM6, LISP, NAT66, and ILNP)?
Best wishes, --Eric -----Original Message----- From: rrg-boun...@irtf.org [mailto:rrg-boun...@irtf.org] On Behalf Of Brian E Carpenter Sent: Thursday, April 15, 2010 2:46 PM To: Templin, Fred L Cc: Tony Li; IRTF Routing RG Subject: Re: [rrg] Proposal for recommendation language Hi Fred, On 2010-04-16 05:00, Templin, Fred L wrote: ... > But, how can we give the customers the perceived advantages of PI > without locking them into a PA "matrix"? > How can we make it so that the ISPs serve the customers instead of the > other way around? The IRON-RANGER solution to this is simple; have the > ISPs give the customers PA addresses, but then let them use the PA > addresses as a springboard into PI. Then, the customer border routers > get exposed to PA addresses, but the customer internal networks get to > use PI and get to multihome whether or not multi-addressing is also > used. So why not have a combined PA/PI scenario instead of strictly > one vs the other? Wouldn't that tend to have some bearing on the RRG > recommendation? Well, to take a few other examples, SHIM6, LISP, NAT66 and ILNP can all be construed to provide the same effect, to a close approximation. (You can quibble, for example SHIM6's ULID is in fact an arbitrary PA address, although it behaves like PI for the host, but broadly, they're all isomorphic). IMHO that's why it's hard to choose. All solutions are equivalent at a sufficient level of abstraction, so the choice depends on details, and we're never going to agree on the details. Brian _______________________________________________ rrg mailing list rrg@irtf.org http://www.irtf.org/mailman/listinfo/rrg _______________________________________________ rrg mailing list rrg@irtf.org http://www.irtf.org/mailman/listinfo/rrg