[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: Resolving geo discussions
On Tuesday, Apr 15, 2003, at 16:45 Europe/London, Brian E Carpenter
One of today's constraints, for example,
is "get rid of the packet as quick as you can, unless it's going
to my own customer" and that has major impact on the BGP4 topology.
Hot-potato routing, which you describe reasonably well in
the quote above, is an artifact of the non-use of MEDs.
This is not a very strong requirement, and mostly exists
because of mid-1990s differences in internal metrics between
Sprintlink and InternetMCI leading to the realization that one
could never really rely upon neighbour A's MEDs making sense
in comparison with neighbour B. Decreasing inter-provider trust
over time also eroded the value of MEDs from other networks,
and ignoring/rewriting them effectively led to mutual hot-potato
routing becoming commonplace.
However, it was never universal, and still isn't. There are networks
which accept MEDs from e.g. customers, just as there was a network
which did static MED-like multi-exit handling in the PR^H^HRADB.
The BGP4 topology (assuming we could agree on a definition) is
not affected by the MED/no-MED decision because of the inability
to compare reliably and in a useful way MEDs offered by different ASes
announcing the same prefix.
I think this quickly gets into Research Group territory.
Unless it gets in the way of actual charter-relevant work, I personally
don't mind this sort of discussion happening on this list instead of on
one within the RRG. The people are going to be the same anyway,
for the most part, and at the moment it doesn't really matter in which
forum clue gets applied, so long as it can be (or does).
But it isn't just Moore's law. It's convergence time. And this is
Frank & I would be thrilled if you would volunteer yourself to manage
where we are still waiting for useable output from the Routing
a subgroup focused on producing exactly this output.
Other applicants also seriously welcome.