On Mon, Nov 15, 2010 at 12:27:56AM +0000, Ed W wrote:
> Yeah, but it's way less than 2^128 which is how everyone is looking at IPV6
Again, look at the link I posted earlier. IPv6 current allocation
practices only make it look like IPv4 at the beginning, assuming
all kind of questionable practices (that people are the primary
driver of address allocations).
> I don't know, but I wouldn't be surprised if larger prefixes are handed
> out to those who ask, eg larger ISPs who in turn are handing out /56s to
ISPs are allocated /32, of which /48 is supposed be handed out
to a customer. I've asked for a /48, but got a /56, which is
not a good policy according to recommendations, since I'm supposed
to issue no less than a /64 to end users, because /64 is a built-in
assumption in a number of current practices which would then break.
> customers. There is a risk of handing out too large a prefix here just
> like in the early days of IPV4 where they handed out /8s to anyone who
> asked...
Exactly.
> 10^7 just doesn't look like much of a safety margin when you can easily
> imagine reducing that by 10x - 1,000x in the near future?
>
> I think the point is that IPv6 /32s are way beyond current router
> capability, so really the standards backed up to something targeting
Anything which produces exponential memory demand (and router
table growth showed that before classless networks became the
norm, and now show it again) in every piece of core infrastructure
is goin to break. The only way to make a global network scale,
especially when cut-through routing is down the corner is to use
geographic purely local-knowledge routing. You can assume that
a few decades hence this will have to be shoehorned into IPv6,
and you can assume that legacy network assigments will be in
the way. The only way to avoid that is to establish a completely
new protocol, and tunnel legacy IPv6 through it.
> more like some million or so times greater than current capacity from a
> pure router capability point of view. This is great, but it's way less
> than the 2^96 fold improvement you might naively expect.
>
> I guess router manufacturers can pull a 10^6 fold performance
> improvement out of somewhere in the next 35 years - question is whether
I don't think that the Internet will look much like current networks
in 35 years.
> demand for zillions of new routed devices will waste the large ipv6
> block allocations rapidly...
-- Eugen* Leitl leitl http://leitl.org ______________________________________________________________ ICBM: 48.07100, 11.36820 http://www.ativel.com http://postbiota.org 8B29F6BE: 099D 78BA 2FD3 B014 B08A 7779 75B0 2443 8B29 F6BEReceived on Mon Nov 15 09:27:32 2010