On 2007-12-20 15:46, Dino Farinacci wrote:
On 2007-12-20 09:32, Dino Farinacci wrote:Ran, first off, really good post.So I would suggest that folks think about IPv4 and IPv6 solution approaches separately. For example, while one might want one of the existing proposal for IPv4 (partly for expediency and partly because IPv4 has more constraints), one might well want a different more architecturally fundamental change for IPv6 (partly because the protocol is more flexible due to extra bits in the header and partly because we have more time to study, prototype, and design a more elegant solution).So let me propose something: 1) For IPv4, use LISP encapsulation as spec'ed in the -05 draft. 2) For IPv6, use header address translation (of the high-order 8-bytes), spec that out as GSE++.As far as I can tell, that whole line of thinking was thrown out of the boat when the RIRs and ISPs declined to accept the recommendation that *all* site prefixes should be /48 or shorter. I don't see any way to revive a (6+2)+8 type of proposal now, unless you can figure out how to deal with variable length routing goop.Actually might be a good thing. Those 2 bytes will be useful (read: necessary) for site subnetting. So maybe we should call it 6+2+8 where the 6 is global-locator, the 2 is a site-based locator, and the 8 is an ID.DinoP.S. The 8+8 proposal didn't clearly tell me how sites wouldn't have to subnet renumber when they gave the Routing Goop back. Again, need the decoupling.
Sure, but that was covered in the 2nd ("GSE") version draft-ietf-ipngwg-gseaddr-00.txt which quickly replaced the original 8+8 draft. The problem for sites arises if different ISPs give them different prefix lengths --- that will really mess things up at the subnet level. On 2007-12-20 15:54, Tony Li wrote:
We've done variable length before... ;-)In any case, if we go down this path, we are making enough other changes (e.g., shifting away from PI prefixes) that we shouldn't feel overly constrained. We could decide, for example, to reserve some address space for 6+10 and then migrate into that...
Sure. There's plenty of space to do that. But don't forget the other big problem of 8+8 or GSE -- the transport layer has to be fixed up too, unless you can arrange for all goops to have the same checksum ;-) Brian -- to unsubscribe send a message to rrg-request@psg.com 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