throughput and tx rate in marginal signal conditions

[Jean Tourrilhes]

don wrote :
> The hostap driver (as I read the code) drops the bit rate each time that
> there is a transmit error and puts it up if 50 good packets are sent. If
> the bit rate setting was based on the errors to good transmit ratio of
> each bit rate, then I think that the throughput under marginal signal
> conditions would improve considerably.

	<Shameless plug>
	You may want to read the paper I wrote titled "Fragment
Adaptive Reduction", it covers some of this subject.
	</Shameless plug>

	Also, I remember that a Bell-Labs paper or Lucent publication
describe the scheme adopted on the Wavelan IEEE card, but I can't find
the reference here. There are also probably lot's of papers on the
subject (use Google).
	It's hard to design an algorithm that cover all possible
scenario, because errors come from various place, so before embarking
on such an adventure, get some info on the subject.

	I agree with you that dropping the rate on a single packet
failure seems a bit exessive, especially if there is contention. As a
simple fix, I personally would vote for two consecutive packet
failures (which if I remember is the Lucent solution).

	Have fun...

	Jean