Saverio, - | we have recently submitted to Infocom 2009 a paper entitled: | | "A Mismatch Controller for Implementing Rate-based Transport Protocols" | | that is definitely related with your email. You can find the paper at the url: | http://c3lab.poliba.it/images/0/06/Rmc_ratebased.pdf | | Basically we show that a simple feedback loop can compensate and | counteract uncertainty in timers duration due to clock granularity, | shared Operating System etc. In this way, the sensitivity to timer | granularity is greatly reduced. Thanks so much for sharing this information. And it is excellent timing -- the algorithm appeared exactly at the right time, when it was needed :) Maybe you have seen the earlier discussion, the background was that we had some consensus in this discussion that simply using high-resolution timers will not improve the situation much for the CCID-3 sending algorithm. I have had a read through the paper - the rate output graphs look great -- no more of the "wild oscillations" of X_recv that one could observe in CCID-3 test runs. And I think that it will integrate very well with the goals outlined in RFC 5348, especially since it defines rather the goals to achieve (a smooth sending rate) rather than prescribing a specific implementation. There is no doubt, this calls for implementation. Details may need some rethinking. I will start at the weekend to tidy up the tree, revising the interface and remove now obsolete high-resolution timer parts that would have lead to a much more complicated implementation. We can then start afresh, with this new algorithm. Thanks a lot Gerrit -- To unsubscribe from this list: 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
