Improving the Robustness of TCP to Non-Congestion Events
draft-ietf-tcpm-tcp-dcr-07

[Ballot comment]
Section 1, page 4, Second paragraph:
As SCTP is mentiond as a potential adaptation. Can NCR also be adapted for DCCP's CCID 2 (TCP like congestion control). If that is the case should some text be added here?

Decided to IETF Last Call this due to the importance of TCP. Chair suggested to add text to the LC anouncement that clarifies that this is meant for experimentation, not deployment.

(Isn't the secretariat supposed to insert this when they go to pub requested?)

PROTO writeup for draft-ietf-tcpm-tcp-dcr-07.txt:
"Improving the Robustness of TCP to Non-Congestion Events"

Ted Faber is the shepherd and WG chair shepherd for this document.  The
document is ready for publication as an experimental RFC in his opinion.

The document has had an adequate review from WG members.  in particular,
this is the result of the combination of two documents addressing the
same issue.  Allman and Blanton had originally written a similar
document and decided that the Bhandarkar/Reddy document was a sounder
idea.  They added their insights and editing to produce the current
document.  As a result both the constituent ideas and the final document
have had significant technical review.

The document is primarily a transport document and an experimental RFC
as well, so I believe that the TCPM technical review has been more than
adequate.

I have no specific issues with the document, nor were significant issues
encountered during WG discussions.

WG consensus, while not rousing, is solid.  There are no major factions
or loud critics of the document.  Input was received from a fair number
of WG members.

No appeals or other obstructions of the document have been threatened.

I have done a full ID-nits pass, including the split of references into
normative/informative.  The document does not depend on any I-Ds
awaiting publication.

Technical Summary

Currently TCP treats the receipt of three duplicate acknowledgments as
an indication that a packet has been lost in the network due to
congestion and slows down.  In networks where packets are reordered
frequently, e.g., wireless networks, that reordering may trigger
spurious slowdowns of TCP connections.  The document proposes delaying
the full congestion response for approximately a round trip time while
using TCP Limited Retransmit[RFC3042] to continue to make progress while
determining the cause of the duplicate ACKs.

WG Summary

Originally two documents addressed this phenomenon, but the two systems
were combined, pretty seamlessly, into this current system.  There was a
healthy amount of review and revision of the document, but no real
friction in achieving consensus.

Protocol Quality

Though the system has been studied in simulation, there are no
implementations as yet.  The document is intended as an experimental RFC
that will encourage experimentation.