Andrea Bittau wrote: > This is an experimental patch to bridge CCID2 and existing tcp congestion contrl > algorithms. Added support for "tcp_congestion_ops" in CCID2. This allows for > CCID2 to use different variants of congestion control. The default is the RFC > CCID2, named with much originality, "ccid2". > > I added support for a "tcp adaptor" which allows existing tcp algorithms to be > used withing CCID2. They may be selected by asking for "tcp_name", where name > is "reno", "vegas", etc. > > The following TCP algorithms are believed to work [at least partially] with > appropriate byte counting turned off: > bic, compound, cubic, highspeed, htcp, reno, scalable, vegas, veno. > These don't seem to work properly [I haven't added full support yet]: > hybla, lp, westwood. > > This patch is mainly an RFC. If people seem to like the idea, I can work on > making the patch look good. Keep on experimenting, this is the spirit, that is why tcp_congestion_ops is in inet_connection_sock, eventually I want to have it renamed to congestion_ops, and your experiments will help in diagnosing whatever change in the existing abstractions is needed for us to get closer to this goal :-) - Arnaldo - : send the line "unsubscribe dccp" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
