Handoff insights with Prism2 linux driver

[Dominique Blas]

Le Vendredi 1 Novembre 2002 18:50, Victor Aleo a ?crit :
> Hi all,

Yep, I think this kind of issues are going to be more and more common with 
time going on.

Indeed we'll progressively need to study and then enable roaming between 
adjacent APs and thus to chase anything that delays reassociation.


Studies have been published about ways to accelerate the sequence :
	looking for a new AP in range - deassociation from current AP - reassociation 
with next AP forward.

These studies make their point to prove that a sort a predictive association 
is needed.
The previous sequence should then be :
	looking for all the AP in range - make a map of them all - association and 
	preauthentication with APs that seem to be in the way the mobile is going to 
(predictive) - deassociation from old AP - fast authentication with the next 
AP forward.

Measures tend to show that with this kind of behaviour you could expect 
handoff delays of 20-50 ms that should be short enough for multimedia
content to play with.

Of course these studies don't take into account the fact that the chipset 
doesn't answer correctly and delays the normal behaviour we ought to expect 
to.

Between 50 ms and over 2000 ms there is plenty of place for improvement !!!
I'm agree with you that such a delay is inadmissible but now we are to find 
what is going wrong : 
	is it the chipset with it's secondary firmware 
or
	is it hostap that need to add/replace a few functions a tertiary firmware is
	 supposed to have ?

Thanks,

db