Hi Michel, On Sat, 2002-11-02 at 15:08, Michel Py wrote: > Mark, > > >>> Mark Smith wrote > >>> Obviously my last two models don't really fit the idea that > >>> site-local addressing is to cover a single geographical site. > > >> Richard Draves wrote: > >> Why do you think that site-local addressing is tied to > >> geography in any way? > > > A few reasons : > > 1) because of the name ?! the word "site" in common English > > has geographical connotations. > > 2) the explanation in the Scoped Addressing Architecture > > RFCs / drafts : draft-ietf-ipngwg-scoping-arch-04.txt > > [snip] > > While I fully agree with your reading of the RFC and the English > connotation of the word "site", it appears to be an unwritten convention > that says that a site is a /48. This leads to some semantic confusion in > the case of organizations that get a /48 and are spanned over several > physical locations that could be separated by hundreds or thousands of > miles.
Agreed. The 54 bits for subnet allocation in a site-local address tends to strongly suggest that a site-local addressing boundary is not a geographical boundary. The huge number of subnet bits also tends to suggest that within the administrative boundaries of a connected network, only one instance of SL addressing should exist. For simplicity reasons, my IPv6 network would have a single instance of site-local addressing, and if I had to merge it with another IPV6 network using site-local addressing, I would be working towards achieving a single site-local addressing instance again, as part of the process of merging the networks. Merging the two networks won't be hassle free, but at least a number of IPv6's features will make it a lot less of a hassle. However, I believe a lot of non-IPv6 ML subscribed people (eg your typical net admin) would take the name "site-local" (and the associated definition) literally, ending up with a lot of separate SL address space instances in their network. They will not like the network addressing complexity IPv6 appeared to have forced upon them. I think there could be value in coming up with a more appropriate name and definition. While the name "site-local" is obviously not the only cause of confusion and concern about site-local addressing, I don't think it helps. If a more expressive name and definition is decided upon, it can also help make a number of related questions easier to answer. For example, people will understand better where multi-site routers typically would or wouldn't be used, which makes questions such as whether multi-site support should be a requirement of a router implementation simpler. Thanks, Mark. -------------------------------------------------------------------- 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] --------------------------------------------------------------------
