> i think that some form
> of cooperative path computation strategy
> ( e.g. diffusive computational nash game, or
> epidemic/gossip tabu search, or whathave you)
Cooperative path computation are highly theoretical at
the present stage. I think there might be many
assumptions and approximations to prove such strategies.
take one simple example: Throup et all describe
methods for optimizing OSPF weights by using heuristic methods
..esp local search hueristics (tabu...) But in reality, people
generally use standard heuristics (like Weight inversely propotional
to capacity.) So i think as a least option we may choose some standard
heuristic assumptions for solving such co-operative path computations.
So using tabu search methods are sometimes not practical.
>might yield
> more novel solutions than incremental techniques rooted in
> the 35 year old stuff we've (oops, s/we/me) tended to get
> stuck on....
Solutions might be novel. We introduced IPv6 long before but still
we could not transfer to new structure all together. we go for
increment methods like CIDR, NAT, double stack, and ALG than transition
to a new addressing structure. So, it is always advisable to do some
incremental techniques than novel solutions. But, if you are proposing
new structure, first we should address flag day problem will
trasitioning. i dont think any new systems does this work.
This archive was generated by hypermail 2b29 : Mon Aug 04 2003 - 04:10:04 EDT