pads & plpmtud BOF

["James Kempf" <kempf@docomolabs-usa.com>]

These were two TSV BOFS, different as night and day.

PADS:
--------
I had a hard time understanding initially what the problem was that these people
wanted to solve. Eventually, during the discussion period, Henning Schulzrinne
stood up and explained it (at least, I think he did, it wasn't apparent that the
chairs shared his view). NSIS has decided on a path coupled approach to
signaling, which I belive means that the signaling and data paths are the same.
But, there are some uses of path decopuled signaling and I think these folks
want to look at that. As Henning explained it, the architecture is the opposite
from RSVP. In RSVP, the control signaling goes through, then the COPS policy
signaling confirms. In  PADS, the policy signaling goes through (possibly
inter-AS) and then the control signaling spreads out. There was also some
confusion about whether routing might be affected by the control signaling.
Henning also explained to me privately that the signaling could be used to set
up authenticated and authorized bandwidth allocations so that the problems with
DoS attacks using end to end QoS could be eliminated.

Presuming I've gotten this right (which, again was hard to tell from the BOF), I
think these folks need at least one more BOF to basically get their act
together. They need to refine their problem statement to make it more precise,
and describe exactly why NSIS isn't the right place to do this.

PLPMTUD
------------
PLPMTUD is a new way of doing path mtu discovery, The standard algorithm is
often error prone, the goal of this working group would be to make MTU discovery
really robust. Implementations of the algorithm exists.

I would recommend that this WG be chartered, since it seems to have  a very
focussed problem statement and clear idea of goals.