Re: PI/metro/geo [Re: The state of IPv6 multihoming development]

[Iljitsch van Beijnum <iljitsch@muada.com>]

On Sat, 2 Nov 2002, RJ Atkinson wrote:

> and some few of us suspect there might be a convergence time wall
> not very far in front of us in IPv4-land, that is not due to "weak
> vendor hardware", but instead due to inherent issues in the current
> Path Vector algorithms behind BGP.  This kind of suspicion is part
> of why the IRTF RRG is examining the broader topic of IP routing --
> though that level of research is outside the charter for this
> list....

Doing this ourselves would be stretch charter-wise, but it would be good
to be aware of what's going on, otherwise we may come up with the
perfect solution that will last three months until BGP goes down the
crapper.

In other words: any information you have on this will be appreciated.

And what I'd like to know: we're rapidly approaching the moment when
routing information about the entire internet is too much to be present
at a single point in space and time, even with provider aggregation in
place. So how will this work with a link state protocol? The advantage
of a distance path protocol is that it also works when the routing
information isn't fully converged.