[Gen-art] Gen-ART telechat review of draft-ietf-tcpm-accecn-reqs-07

[Brian E Carpenter <[e-mail]>]

I am the assigned Gen-ART reviewer for this draft. For background on
Gen-ART, please see the FAQ at
<http://wiki.tools.ietf.org/area/gen/trac/wiki/GenArtfaq>.

Please wait for direction from your document shepherd
or AD before posting a new version of the draft.

Document: draft-ietf-tcpm-accecn-reqs-07.txt
Reviewer: Brian Carpenter
Review Date: 2015-02-13
IETF LC End Date: 2014-12-04
IESG Telechat date: 2015-02-19

Summary: Almost ready
--------

Comment:
--------

This is a well written document. However, I saw no response to my LC
review (2014-11-25) so it is repeated below.

Minor issues:
-------------

There's quite a lot of discussion of the issues that would be caused by
lost ACKs, but it's also stated that "(in the worst case, loss will still
be available as a congestion signal of last resort)" and "However, it
should be noted that ECN feedback is not the last resort against
congestion collapse, because if there is insufficient response to
ECN, loss will ensue, and TCP will still react appropriately to loss."

This doesn't address the issue that on physically lossy networks
(e.g. the networks that more and more user devices live on), TCP does
*not* react appropriately to loss, because it treats it as a congestion
signal, and slows down when that is completely the wrong thing to do.

I think that the draft should recognise the fact that when a physically
lossy network is involved, ACK loss will be a real issue at exactly
the same time that conventional TCP is liable to misdiagnose congestion.