[HECnet] Network topology according to MIM.

[Robert Armstrong]

>you (I think) and Peter have been believing that you have a problem that
> can be solved by playing with costs in a way that will not solve the
problem.

  I can't speak for Peter, but all I want from the circuit costs is two
things -

  * The routing should be symmetrical; that is, if you ask node A for the
path to node B, the result should be the same as if you ask B for the path
to A.  This isn't an absolute DECnet requirement, but it makes network
problems so much easier to figure out when data is taking the same path in
both directions.  Besides, in HECnet I can't really think of a good reason
to prefer an asymmetrical route, so when you find one it almost always means
something is configured wrong somewhere.

  * If two nodes, A and B, have a direct point to point link (probably via
Multinet, but maybe something else) then that should be the preferred route
between those two nodes.  Other nodes besides A and B are another matter,
but two nodes that are directly linked should use that link.  This seems
obvious on the surface - if the owners of A and B didn't want that result,
why set up the direct link in the first place?  Besides, any esoteric
network principles aside, common sense says that a direct path between two
nodes is faster and more reliable than a link that goes thru a third party,
e.g. psilo.  

>You are essentially hitting up against a problem with DECnet routing, 

  If we were consistent across HECnet about setting network costs, then I'm
pretty sure those two things could be accomplished.

Bob