> what is your position on probably unique local addresses versus > site-locals? as probably unique addresses have a limited topological > span wouldn't applications still have pretty much the same issues as > with site-locals?
no. first, probably unique addresses don't have any strict limits on their topological span - done right they could be routed to thousands of networks - just not to *every* network. second, the biggest problems with site-locals are not that they might get exposed to areas of the network from which those addresses are not reachable, but that they're ambiguous. 'probably unique' addresses (given a sufficient number of random bits) are effectively unambiguous. it seems like the goal here is to find the best mechanism to encourage the sites that use these addresses to pick prefixes that are in practice unique within the transitive closure of the set of networks to which they connect. this is mostly a human factors question. right now I'm thinking that the real trick is getting network operators to learn how to get a (probably or not) unique prefix by any means - and that the difference between going to a registry, running a random number generator, deriving a prefix from some other number, or getting a prefix from a tag stuck to a router - is probably noise compared with the educational effort common to all of these. Keith -------------------------------------------------------------------- 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] --------------------------------------------------------------------
