Note: This ballot was opened for revision 09 and is now closed.
Summary: Needs a YES.
Clearing my Discuss now I have found Section 12.3.
Can I suggest that you rebrand sections 12 and 13 not as appendixes. Also that
you so not use the first person plural in the text of secition 12. I think the
RFC is supposed to represent at least WG consensus and so the first person is a
This idea appears to have the issue that if the variance introduced by new
sessions coming on exceeds the tolerance in the marking zone (eg, if there is
plenty of capacity for one HDTV stream but not enough for two, and someone
turns on a second TV while the first is watching a show), that can impact the
application in the existing channel. The simulations have tested the ability of
the system to operate when properly operated and properly configured; they have
not seriously considered avalanche scenarios.
Avalanche scenarios are a variation of the "mother's day" problem, on a
millisecond time scale; in real time applications such as voice and especially
video it is possible to outrun an essentially infinite bandwidth pool during
short time scales. Consider concerns with realtime video conferencing and the
very careful planning to ensure that the network can handle that class of
traffic - on a 45 MBPS link with three data streams that nominally run 5 MBPS
standing and ~12 MBPS at peaks (I-frame). There are demonstrated issues in
picture quality due to conniption fits in the network and as a result force
more careful traffic pacing.
I think this is fine in parts of the network where avalanche scenarios are
unusual but, that has to be clearly stated and that the idea does not work in
other use cases.
The Gen-ART Review by Francis Dupont on 2009-02-23 provides many
minor suggestions. Please review them.