[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: BCP for multisite multihoming
Hi,
On Mon, May 21, 2007 at 04:35:20PM -0400, Leo Vegoda wrote:
> Yes, a /32 is far more than many sites will need.
>
> The first requires a new, scalable approach to routing and the latter
> can be solved by changing the minimum allocation size. The latter is
> relatively easy.
Technically, this is the wrong forum, but I want to make this heard
anyway. I consider every thought of reducing the /32 to a smaller
minimum allocation size fundamentally flawed - this is IPv4 thinking.
IPv6 allocation policies should aim at
- a single enterprise should be happy with a *single* IPv6 prefix,
for as long as the network exists (so it needs to be big enough for
your typical small-to-medium sized ISP - and a /32 just meets that)
- it is not much use to squeeze out address space on things that need to
go to the routing table anyway (ISP prefixes, that is) - what do you win
if you force /40s on ISPs, and then end up with 99% of the address space
unused because your routing table cannot take more entries?
Giving /32s to ISPs means, theoretically, 500 million /32s out there -
which everybody agrees that routers cannot take. So, reducing this to
"smaller blocks", whill indeed save space - but what can you use it for?
The /32s are already a compromise, and as soon as an ISP has an internal
network hierarchy, needs to do internal aggregation etc., this isn't so
spacious anymore. I would have preferred a /28 or similar "no questions
asked".
Please stop thinking in IPv4 trains.
The real question is "who needs/deserves a global routing table slot",
not "how many bits can you haggle for".
Gert Doering
-- RIPE APWG chair
--
Total number of prefixes smaller than registry allocations: 113403
SpaceNet AG Vorstand: Sebastian v. Bomhard
Joseph-Dollinger-Bogen 14 Aufsichtsratsvors.: A. Grundner-Culemann
D-80807 Muenchen HRB: 136055 (AG Muenchen)
Tel: +49 (89) 32356-444 USt-IdNr.: DE813185279