12 Nov
1996
12 Nov
'96
9:31 a.m.
Quick comment: You're looking at the computing of routes wrt one router. I'm suggesting that perhaps we should consider the set of all routers since we're concerned about spreading load due to the quick communication from one router to many of every best external-path changing.
Some people are concerned with that and others are concerned with end-node (or close to it) routers where this is not an issue. I suspect nanog has a good mix with many who run both types.