Dag wrote :
> Jean Tourrilhes <[EMAIL PROTECTED]> writes:
> 
> >       I believe that having many speed doesn't matter much, because
> > 115kb/s is almost as reliable as 9.6kb/s, so as long as you do 9.6kb/s
> > and 115kb/s you are ok.
> 
> Not with the Palm Pilot (see below)

        Point taken, thanks...

> Well, here is some thoughts about dongle speed:
> 
> 1) Even though a dongle can change speed fast, the peer device may still
> use 200 milliseconds or more to change its speed. So changing the speed
> fast may not help anything. So if you change the speed fast, then the peer
> device may miss the first RR frame, which means that you must retry after
> some timeout which will make it all take _much_ longer time, than if you
> had used longer time to change speed ;-)

        I know that IrDA negociates the turnaround between the two
device, choosing the lowest of the two values, and I was hopping it
was dong the same for speed change. Anyway, the IrDA standard must
provide some delay there, but I just feel that the dongle should not
add to that delay...
        In your case, you would need both sides to be fast. If the
other side is the same dongle, then I win. I don't know what most
laptop implementation does, but I guess that if it support 4Mb/s it
must have a good turnaround time...

> 2) Minimum turnaround time has a big impact on performance, and the

        That's why I was complaining about the values found in the esi
and tekram drivers. Maybe the doc would show some better values than
that...

> 3) The JetEye is probably the most stable dongle because the impl. is so
> simple,

        Yes, amazingly simple... But wait a second for the old_belkin...

> but the JetEye doesn't support 57600 which is used by the Palm
> Pilot, so it you use the JetEye with the Palm you will only get 19200
> bps. One other thing with this dongle is that the cable is not very well
> attached to the dongle, so the cable will slip out after some weeks and
> expose the wires inside :-(

        But, it has a nice little light on top (which could be put to
good use but as it is it is useless). And also turnaround time is large.

> 4) The Girbil dosn't do echo-cancellation which means that you must use
> irport with this dongle. The girbil is however the nicest and smallest
> dongle ;-) I have had some problems with my impl. of this dongle, but they
> may be fixed now in the new dongle arch.

        But difficult to find, it seems...

> 5) The Tekram dongle is quite large, but it's working very nice (at least
> with the patches I made today ;-) It supports most speeds, and also
> TV-remote mode I think.

        Turnaround is large as well.

> 6) The LiteLink dongle is quite big, but the impl. is also quite simple so
> it should be stable to use (at least with the very latest patches ;-)

        Yep.

> 7) Although I have written the driver for the ACTiSYS dongle, I don't have
> it :-( So I cannot tell you anything about this dongle.

        If the person from Actisys confirm the numbers for speed
change, then the driver could be made as simple as the litelink.

> Hope this will make it easier for you to choose a dongle. 

        Yep.

> DISCLAIMER: Any dongle problems listed above may be problems with
> Linux-IrDA and not with the dongle itself. I have just tried to tell you
> how this dongle works with Linux-IrDA (it might work better with Windows).

        Moreover, everybody has it's own criteria and requirements, so
best dongle is only with regards to specific requirements ;-)

> -- Dag

        Jean


_______________________________________________
Linux-IrDA mailing list  -  [EMAIL PROTECTED]
http://www4.pasta.cs.UiT.No/mailman/listinfo/linux-irda

Reply via email to