----- Original Message ----- From: "Brian Haberman" <[EMAIL PROTECTED]> To: "Stephen Sprunk" <[EMAIL PROTECTED]> Cc: "Bob Hinden" <[EMAIL PROTECTED]>; "Mark Smith" <[EMAIL PROTECTED]>; <ipv6@ietf.org> Sent: Monday, January 17, 2005 14:49 Subject: Re: Proposed update to ULA Draft (-09)
> > On Jan 17, 2005, at 14:57, Stephen Sprunk wrote: > > > Thus spake "Bob Hinden" <[EMAIL PROTECTED]> > >> At 04:40 AM 01/17/2005, Mark Smith wrote: > >>> (as a side note, this is from Rev 8, the nokia web site resolves to > >>> an > >>> IPv6 address, I don't seem to be able to get to it via my 6to4 > >>> connection though) > >>> > >>> For link-state IGPs, it is suggested that a site utilizing ULA > >>> prefixes be contained either within one IGP domain or area. By > >>> containing a ULA prefix to a single link-state area or domain, the > >>> distribution of prefixes can be controlled. > >>> > >>> I think it potentially could cause conflicts with the idea of using > >>> the > >>> same subnet numbers as those used with a global prefix. > >> > >> The language "it is suggested" was chosen to allow flexibility. As > >> you point out in large multi domain/area deployments there are > >> reasons to do it differently. > > > > I think removing "area" and just leaving "domain" makes more > > sense; I can't see anyone actually deploying different ULA prefixes > > per area. The only possible justification would be different ULAs > > per business unit (to prepare for sell-offs) but business unit lines > > are rarely reflected in topology, in my experience. > > This change was made to address a DISCUSS comment specifically > talking about link-state routing protocols and area boundaries. The > text of the comment is available at > https://datatracker.ietf.org/public/pidtracker.cgi? > command=view_comment&id=26775. I can't find a way to agree with Alex's comments. A "site" certainly can't be (productively) smaller than a link-state area due to flooding, but it can certainly be bigger than an area and even span domains if they're under the same administrative control. Part of the problem is "site" has been overloaded to mean "administrative domain", whereas the common meaning is a single physical location. Neither makes sense to me in the 09 text, though. > > Also, why does this suggestion only apply to link-state IGPs? > > What are our recommendations for people using RIPv6, EIGRPv6, > > etc? Or with internal BGP peering between multiple IGPs? > > Distance-vector protocols are not generally broken up into admin areas > like OSPF & IS-IS. EIGRP is often broken up into "areas" of a sort due to summarization, and the consequent limitation of routing updates appears similar to link-state protocols. With EIGRP v1.1, each physical site may end up being a "stub" area; should every location then get its own ULA prefix, even if it only has one router? > I used to run a modified ripv6 daemon that routed the original > site-locals without issues (wrt internal routing). So, I don't see > much need in giving a bunch of guidance there. > > As to internal BGP, my question is "How many scenarios do we > want to carry around in this document? I question whether we should carry any. The only strong requirement is that a "site" be contained within a single AS, regardless of how many routing protocols or areas it contains. Network engineering dictates how many prefixes will be required in a given site's configuration. S Stephen Sprunk "Stupid people surround themselves with smart CCIE #3723 people. Smart people surround themselves with K5SSS smart people who disagree with them." --Aaron Sorkin -------------------------------------------------------------------- IETF IPv6 working group mailing list ipv6@ietf.org Administrative Requests: https://www1.ietf.org/mailman/listinfo/ipv6 --------------------------------------------------------------------
