[ppml] Solicing comments: IPv4 to IPv6 Migration IncentiveAddress Space
jeroen at unfix.org
Wed Jun 27 11:03:52 EDT 2007
[I also *FAR* from like this proposed proposal, if folks want IPv6 let
them get it from the RIR, it is there they just need to ask for it]
William Herrin wrote:
> The question was: how would this proposal impact the routing table. In
> the worst case scenario, this proposal would place the same number of
> routes in the IPv6 table that are presently in the IPv4 table: roughly
No, you are looking at it from the wrong way. At the moment there are
about 40k active ASN's who indeed announce about 200k IPv4 routes.
In IPv6 there are about 1000 prefixes (of which ~800 good ones) in the
IPv6 tables, with about 800 ASN's announcing them. That is a ~13%
overhead. Thus if every internet-active ASN would announce a single
prefix + the overhead you would get 40k + 13.3% is only ~55k prefixes.
But we have ASN32 nowadays so that can explode up to 4 million ;)
> The amount of memory necessary to list every possible subnet in IPv6
> is not at issue and not relevant to the discussion.
> On the other hand, my math was in error. Because the IPv6 subnet size
> is fixed at /64 by the standard, routers need never consider or store
> the last 8 bytes of the route. Accordingly, the IPv6 route table in
> the worst case scenario need only consume twice the memory of the IPv4
> table, not 4 times as much.
You are wrong again though, routers need to handle the FULL 128bits. And
fortunately currently they still do that.
It might be one day though that routers might have a fast-path handling
only the first 48 bits and a slow path for handling the rest.
Currently that is not the case and the IPv6 RFC's also state that
hardware and software must route on the FULL 128bits. How they do that
is their problem of course.
-------------- next part --------------
A non-text attachment was scrubbed...
Size: 311 bytes
Desc: OpenPGP digital signature
More information about the ARIN-PPML