[lug] Linux and wireless

[Nate Duehr]

Jonathan Corbet wrote:
> Craig Talbert <Talbert at colorado.edu> wrote:
> 
>> AFAIK the ndiswrapper works for most popular wireless devices:
>>
>> http://ndiswrapper.sourceforge.net/
> 
> I would avoid that if at all possible.  You're putting non-free software
> into your free kernel.  Ndiswrapper will break with each kernel upgrade
> and need fixing - and, for distributions which ship 4K stack kernels
> (Fedora) there *is* no fix.  No kernel developer will help you out if
> you have troubles.

Rebuild the kernel with the other stack size?  What's the big deal with 
that?  Does RHEL not offer that option in their kernel source?

> Linux wireless is at a bit of a low point; see, for example,
> http://lwn.net/Articles/167270/.  But it is getting better in a hurry;
> by the end of this year, I expect the situation to be much improved.
> Watch LWN for coverage from the wireless summit in April :)

Low point?  When was it ever good?  Perhaps back when the original 
Orinoco Gold Card and Prism chipsets were popular, three years ago? 
It's been on a long downhill slide, driven by clueless manufacturers.

The main reason to avoid ndiswrapper is to financially penalize the bad 
manufacturers who don't release drivers for all OS's.  But they don't 
really appear to care, anyway...

> That doesn't help you right now, though.  For the short term, look for
> the Intel IPW chipsets; they are well supported in current kernels.  For
> Atheros chipsets you can use the highly experimental native driver from
> www.ath-driver.org, or go with the madwifi drivers (madwifi.net).  Some
> distributions (i.e. Ubuntu) will install madwifi for you.  Madwifi still
> has a proprietary blob, but at least it's a meant-for-Linux proprietary
> blob.  For Broadcom chipsets, there is an experimental driver in
> reasonably good shape: http://lwn.net/Articles/162894/.  Between all of
> those, there's actually a fair number of options out there.

But very difficult to match up Vendor names with chipsets.  Standing in 
a typical computer store, staring at boxes, noticing that there's now a 
new "rev 04" of some stupid card that had the right chipset in it three 
revisions ago, and that's the latest information on the web... is a 
maddening experience.

Nate