Michel Py wrote: > Margaret, > >> Michel Py wrote: >... > > Possibilities for how to allocate aggregable > > provider-independent addresses could include (among other options) > > the geographical addresses that Tony Hain has proposed. > > > What term should I use for that? > > If it is any different than the PI we have today, I don't > know, because it does not exist. As I said before, xxPI, or > "something that provides the perks of PI and is scalable". I > guess there is no universal name until you actually see the solution. > > Tony's drafts has to components: > a) An address allocation scheme, based on GPS coordinates. > b) An aggregation mechanism, based on a form of exchange aggregation. > > For the a) part we have a somehow simililar (in the sense > that it's based on geography also) allocation scheme, we > called it GAPI for Geographically Aggregatable Provider > Independent. > http://www.ietf.org/internet-drafts/draft-py-multi6-gapi-00.txt > > I don't like too much the association with the negative connotations > of "PI" (they're about as bad as "NAT"), but that's how we named it. > > For the b) part we have different schemes, which is why we have chosen
> to separate the naming of the drafts. > > > That being said, there is nothing that says that the aggregation > mechanism should be based on geography, although we all seem to go in > the same direction. There is nothing that says that the scalability > should come from aggregation either. > > Technically, I still consider Tony's draft a PI draft, same as our > GFN: > - The basic mechanism to inject the prefix in the routing system > is still there. > - Aggregation is optional as there is no coupling between the > allocation scheme and the aggregation mechanism. > > As I have said before, geographically aggregatable schemes are > superior to PI in any situation, but they are far from being enough alone. Leaving the semantics discussion about PI aside, Michel points out there are multiple ways to handle allocations that are not based on provider hierarchy. They have different aggregation trade-offs, and none of them are going to result in an absolutely optimized routing table for global distribution. The point of aggregation is to mask details where they are irrelevant. In a way, this discussion becomes a debate about the definition and scope of the DFZ. Today the DFZ is expected to be propagated globally. Maybe we should be questioning that assumption. Even if you disagree with the geo approach, I would appreciate comments on the current update to the companion document that discusses the need for PI. http://www.tndh.net/~tony/ietf/ipv6piaddressusage-04.txt I believe the motivations apply no matter which 'PI' mechanism we end up with. Tony -------------------------------------------------------------------- IETF IPng Working Group Mailing List IPng Home Page: http://playground.sun.com/ipng FTP archive: ftp://playground.sun.com/pub/ipng Direct all administrative requests to [EMAIL PROTECTED] --------------------------------------------------------------------
