% Bill Manning <[EMAIL PROTECTED]> writes:
%
% So how many /20s or how many /116s will fit into a routing table?
%
% Alternatively -- if we take Harald Alvestrand's numbers,
% and assume everyone who wants one gets an IPv6 /48,
% leaving 16 bits for subnetting before the "flat space" --
% can we try to quantify the consumption of critical
% resources in routers with this as an explicit goal?
%
% (I think it's fair to make assumptions that we will
% continue with BGP and the iBGP hack, and known mitigations
% of the n**2 iBGP problem. I think it'd also be
% fair to consider time/space tradeoffs in the BGP
% data structures and the data structures associated with
% the actual forwarding of packets.)
%
% Sean.
Raw numbers of prefxies are pretty impressive. A IPv4 /20 is 4k "host"
addresses. An IPv6 /116 is the same sized "chunk" so, the total number of
/20s in the IPv4 world: 1024000 (stuff that into your router)
/116s in a /48 chunk: 34359738368 (is that right?)
Haralds numbers are actually pretty easy. 16 bits of delegtion. Seems
like a routing table that would roughly be the size of a IPv4's /16
in host routes or 'prox what we have in todays table. Minus the transition
routes of course, and the more specifics, and the odd bits on the edge that
don't quite fit the model, and ...
--bill
