On 9/3/08 4:15 PM, Tony Li allegedly wrote: > When Noel was talking about PI addressing, I thought that we were talking > about continuing the legacy process of handing out true addresses. > Obviously, if we change to a locator/identifier solution, then we replace > those with identifiers. Let's *not* confuse the issue by mixing PI and > identifiers. They are completely different.
OK, I now understand that when you say PI you mean a globally routed, non-aggregated address prefix. Yes? However, I think the essential point here is not "PI" (provider independence). Rather, it's the trade-off between non-aggregation and being globally routed. One or the other needs to be controlled by some means. Even PA prefixes should either be aggregated or have their routing constrained. So it comes down to aggregation versus scope: limiting one or the other solves the problem. We can look at the different classes of approach to limiting one or the other, to see which one causes the least pain. I believe that anything that requires *site* renumbering in order to switch upstream providers is going to be hard to swallow. That includes any approach that uses PA addresses within a site, i.e. any approach that aggregates instead of limiting scope, and where an endpoint knows an external prefix for itself. At first glance that would mean at least Handley/Trilogy/Multipath (unless modified) and the IPv6 loc/id split ones except GSE where the "routing goop" in incoming packets is zeroed out at the network edge. It may be that there is pain associated with scope-limiting approaches that is even worse than the pain of site renumbering. Once you have decided to limit routing scope, there is no reason for your site-internal addressing not to be "provider-independent" (strict meaning). swb -- to unsubscribe send a message to [EMAIL PROTECTED] with the word 'unsubscribe' in a single line as the message text body. archive: <http://psg.com/lists/rrg/> & ftp://psg.com/pub/lists/rrg
