[e2e] RFC 1323 revision

Injong Rhee rhee at eos.ncsu.edu
Thu Aug 14 18:38:19 PDT 2003

I am following up on the Reiner's idea on echoing TsRecent. I think it is a
good idea and needs to be supported.

 I believe that there is more gain by allowing TsRecent to be echoed back
for DUPACK than loss. As Reiner stated, if the (modified) sender does not
take RTT samples during recovery, it would lead to more serious problems of
underestimating RTTs. In some networks such as wireless networks, RTT can
change fairly rapidly, especially during loss periods (this may or may not
be due to congestion). If the RTT samples cannot be made during these
periods, the sender ends up getting widely wrong RTT values. Reordering is a
much more rare event than recovery. I wonder how much underestimating would
be possible due to reordering (given that the reordering delays is not much
and also the RTT samples are averaged). Furthermore, if we echo the
timestamp of the most recently received packets instead of the most recent
timestamp (as Reiner pointed out in the earlier email), RTT underestimation
due to reordering would not occur.

Another related issue is that echoing back most recent TS samples can allow
more information inference about the receiver's states for the sender. For
instance, it can lead to better estimation of PIPE in TCP-SACK, and also can
be used to avoid false fast retransmits. Also it can also help in detecting
losses of retransmitted packets (these are not part of standards yet) I have
been pondering on these issues - just been lazy in writing up on them. It
turns out that echoing TsRecent could result in much improved performance
for TCP under high-speed, long distance networks. I will try to make the
paper available soon.

So stay tuned.


Injong Rhee

-----Original Message-----
From: end2end-interest-admin at postel.org
[mailto:end2end-interest-admin at postel.org]On Behalf Of David Borman
Sent: Thursday, August 14, 2003 10:27 PM
To: end2end-interest at postel.org; tsvwg at ietf.org
Subject: [e2e] RFC 1323 revision


A revision to RFC 1323, TCP Extensions for High Performance, is
currently being discussed on the tsvwg at ietf.org mailing list.
Is this the right place for this discussion, or should it move
over to end2end-interest at postel.org, the End to End mailing list?
(The TCP Large Windows mailing list is no longer active.)

                        -David Borman

Incoming mail is certified Virus Free.
Checked by AVG anti-virus system (http://www.grisoft.com).
Version: 6.0.507 / Virus Database: 304 - Release Date: 8/4/2003

Outgoing mail is certified Virus Free.
Checked by AVG anti-virus system (http://www.grisoft.com).
Version: 6.0.507 / Virus Database: 304 - Release Date: 8/4/2003

More information about the end2end-interest mailing list