My slides for tomorrow, this time with the correct URLs...
TFRC-SP:
http://www.icir.org/floyd/talks/voip-Nov06.ppt
http://www.icir.org/floyd/talks/voip-Nov06.pdf
CCID4:
http://www.icir.org/floyd/talks/ccid4-Nov06.ppt
http://www.icir.org/floyd/talks/ccid4-Nov06.pdf
RFC3448bis:
http://www.icir.org/floyd/talks/tfrcbis-Nov06.ppt
http://www.icir.org/floyd/talks/tfrcbis-Nov06.pdf
If there was time, I would also like five minutes to ask some questions
about CCID 2:
* Would people be interested in an option for CCID2 for a
slowly-responding version,
decreasing to 7/8-ths instead of 1/2 in the decrease of AIMD, and
adjusting the increase
appropriately? That is, AIMD, but with different increase and
decrease parameters.
* If people were interested, would they be interested in a
small-packets variant of CCID2?
With at least 10 ms between packets, but using the AIMD parameters
that would
be used with 1460-byte segments, to get the same throughput in bytes
per second as
a large-packet TCP that increases its window by 1460-bytes per
round-trip time?
(The small-packets variant would take header bandwidth into account
as is done for TFRC-SP.)
(People seem to equate DCCP with CCID 3 and TFRC, but they aren't the
same thing...)
- Sally
http://www.icir.org/floyd/
