Stewart & Mike,
In general, I think that the idea of clearly specifying the
applicability is good. What I think would also still be good to
capture is the set of cases that are NOT applicable. This is within
the context of the IGP as well as, possibly, without. For instance,
traffic micro-loops to a new prefix will not be handled. Similarly,
clarifying when & whether the single failure assumption needs to apply
& how this interacts with the multiple areas would be useful.
The interactions of other protocols with the IGP may be usefully
discussed as well. For instance, the IGP controlled convergence may
prevent (or reduce) micro-loops for BGP destinations as long as BGP
itself doesn't need to make a new policy decision.
On a slightly more detailed note, in the draft, you discuss the
possible use of the incremental cost update mechanism for the
management withdrawal of a component. I'm curious why you thought
this was more applicable here. One of the issues with the incremental
cost update mechanism was how long it left the network vulnerable to
another change. I don't see that the cause for that change reduces
the problems with a long vulnerability window.
Rtgwg mailing list