On Sat, 29 May 2004, Simon Leinen wrote:
Short version: Let's ship it!
Long version:
I think the current version is Good Enough, and it's high time we get this out as an RFC, so that the routing registries implement it and we (ISPs that have peerings other than IPv4 unicast) can use it.
Personally I'm not convinced that the current defaulting semantics in the "mp-..." attributes are optimal, but they seem to be close to the best compromise that we can achieve. My reservation is that we default mp-... without "afi" qualifier to default to exactly the four AFI/SAFIs of ipv4.unicast, ipv4.multicast, ipv6.unicast, ipv6.multicast, and this set of four AFI/SAFIs doesn't seem to be guaranteed future-proof. But I don't care, I don't think I'll be using the defaults anyway (look at AS559 in the RIPE test whois registry at rpslng.ripe.net, port 53001, for how I intend to use RPSLng). By the time people will want different defaults - either because IPv4 or IPv6 or multicast (re-)become exotic, or something else becomes widespread - we simply won't be able to seamlessly change those defaults. But probably by that time it will be a good idea to base the routing registry on something completely different anyway.
A unified grammar (maybe even in RFC2234-compatible ABNF!) would be great, but let's leave this for a future RFC that replaces both RFC 2622 and the RPSLng RFC.
Regards, -- Simon.
Hello, I strongly agree with Simon. This is already suitable enough to be taken to the next step (the deployment that will enable non-v4-unicast-only ISPs to express their own routing policies). Regards, ./Carlos -------------- IPv6 -> http://www.ip6.fccn.pt Wide Area Network Workgroup, CMF8-RIPE, CF596-ARIN FCCN - Fundacao para a Computacao Cientifica Nacional http://www.fccn.pt "Internet is just routes (135072/470), naming (millions) and... people!"