[rbridge] draft-ietf-trill-prob-00.txt comment

[Eastlake III Donald-LDE008]

Here is one additional comment from me as a working group member on
draft-ietf-trill-prob-00.txt:

Abstract and Section 1:
	The abstract mentions in order, as problem of bridges, the need
to avoid loops, inability to use alternative paths, and convergence
time. Section 1 mentions in order conversion time and the concentration
of traffic on a spanning tree and resulting bandwidth limitation.
	I have no problem with mentioning all of these but it seems to
me that, in practice, bandwidth/efficiency and, to a lesser extent,
latency due to suboptimal paths are the problems that people are worried
about. Loop safety is important and there have and can be loops but as a
practical matter, they seem to be extremely rare. As for convergence,
I've heard it claimed that RSTP can converge in three times the speed of
light delay across the diameter of the network. Regardless of that
claim, is there really a strong case that link-state convergence will
necessarily be faster?
	While they are certainly of importance, why are we putting
convergence and loop robustness first in these introductory sections
when, as a practical matter, the most notable improvement with TRILL
seems like it will be more efficient use of physical links? Admittedly,
when we get into the body of the draft, "2.1 Inefficient Paths" does
come first....

Donald