At 12:57 AM 9/23/00 -0600, mccreary@colorado.edu wrote:
> 1. What engineering practices are producing the increased growth
> rate?
Just a guess, but I suspect that the increase in punch-through routes
(routes previously aggregated but now punched through other ISPs in order
to obtain better connectivity) is a big part of this. You can look at the
number of routes with long prefixes (available at Geoff's site if not the
others) on that.
I find myself wondering, and have kicked around some notions, whether it
would be of value to add a pair of attributes to BGP and RPSL. These would
estimate the mean TCP RTT and mean variance on the TCP RTT between here and
there by summing the values along a route; somewhere along the line, there
would need to be a ping or something to get a first cut value as -
especially initially - there is no obvious reason to believe that everyone
would be capable of it or would generate it. RPSL policies could be written
that said "I am willing to go to here via this that or the other AS, and
let's take the one with the least RTT" or "...the least variance".
No response from a service provider yet saying whether they would like that
or not. I'm not going to pus it if nobody wants it, but it seems like a
reasonable thing.
This archive was generated by hypermail 2b29 : Mon Aug 04 2003 - 04:10:03 EDT