[PLUG] Wifi card

[Russell Senior]

>>>>> "Nick" == Nick Aubert <nickaubert at gmail.com> writes:

Nick> Linux support for Atheros chipsets is pretty good.  Two years
Nick> ago I had to do quite a bit of work to get the madwifi drivers
Nick> working on my laptop.  A week ago I set up a Gentoo laptop with
Nick> an Atheros wifi card, and it was a snap, even with WPA
Nick> encryption via wpa_supplicant.  I don't know how Knoppix does
Nick> with madwifi, but kernel support is definitely there.

Madwifi has indeed been one of the more solid linux drivers.  But
considering that none of them has been *that* solid, it might fall
into the category of "faint praise".  Historically, the hostap-driver
for prism chips was really solid.  Some of the other 11b devices had
decent if divergent support.  I have been using madwifi-old for a
couple years with the atheros radio in my thinkpad (I am sticking with
madwifi-old because it works reliably and since I am often diagnosing
weirdness at PTP nodes, I would rather not have that muddled up by
having weirdness on my own radio.  I've used madwifi-ng extensively in
embedded applications, which has been less solid for me.  Madwifi
(both -old and -ng) relies on a binary blob, the so-called HAL
(hardware abstraction layer).  Over the last year to 18 months there
has been a lot of work on the "mac80211" wifi stack for linux, and in
in the last few months also the ath5k driver for atheros radios.  I
haven't tried it yet, but my impression from reading the relevant
mailing lists is that it's still a bit rough.  Which is completely
understandable, considering the lack of support that most driver
writers get from hardware vendors.  My guess is that it'll be another
6 months before things really start converging on "solid" and wifi
support gets better on linux.  In the meantime, find whatever life
raft you can.  "There's hope.  No Luke Skywalker in sight."

  <http://lkml.org/lkml/2006/1/5/671>


-- 
Russell Senior, Secretary
russell at personaltelco.net