Avri Doria:
> One of the issues I find myself bumping into is the difficulty folks
> have in configuring the policies in a network in a consistent and
> tractable way. I think this is partly due to the fact that there
> isn't a standard model for describing the policies between ASs. There
> is certainly a language, but no consistent model for the policies to
> be based on. Would it be reasonable for the irtf-rr to include a
> sub-group working on this sort of issue? I.e. a coherent model of
> constraint policies and the networks they create.
While I think improving policy description would be nice, wouldn't it
make more sense to do away with as much of the hands-on configuration
as possible? I'm not talking about making tools to configure routers
from a model, but rather a routing technology that doesn't require
constant human intervention.
There are aut-num objects in the RADB with hundreds of 'import' attributes
(check out AS4513). It's a lot to expect a human to have read these and
verified them. And if this work is being programmaticly now, as surely
it is, then why not include that as part of the routing technology
itself?
But maybe this belongs in the IWTF (Internet Wishful Thinking Forum).
It's like HAL said: "I've seen this sort of thing before, Dave, and it
has always been attributable to human error."
Still reading RFC 1992,
Shane
This archive was generated by hypermail 2b29 : Mon Aug 04 2003 - 04:10:03 EDT