[Gllug] Wifi USB that works with Linux & I can buy online

Mike Brodbelt mike at coruscant.demon.co.uk
Tue Oct 4 20:56:46 UTC 2005


On Tue, 2005-10-04 at 21:25 +0100, Peter Grandi wrote:

> mike> It's worth noting that the madwifi drivers, while native,
> mike> are not fully free software, and taint the kernel.
> 
> Then essentially all WiFi drivers are not fully free software,
> as the chipset firmware is a rule closed source and proprietary,
> and must be obtained separately, and without that the card does
> not work at all.

You can apply that by extension to much hardware available today.
Personally, I still think there is an important distinction between a
blob of binary code which gets uploaded to a peripheral, and a blob of
binary code which gets incorporated in your running kernel.

Until we get to a stage where you can run a system with fully free
hardware, there is always going to be non-free code around in your
system, whether as firmware in an eeprom, OS uploaded firmware, or
etched in silicon (though the definition of "code" gets stretched here).
Nevertheless, I think it's still a worthwhile objective to run a fully
free OS, and I disagree strongly with the madwifi developers on this,
who make the argument that there's no difference between a binary only
module running on the host CPU, and one that gets uploaded to a
peripheral. If one swallows that argument, you might as well not have
free drivers at all.

> The specific issue with the 'madwifi' driver is that part of the
> firmware is bodily incorporated into the driver, instead of a
> clean separation between driver and firmware as in other WiFi
> drivers...
> 
> The OpenBSD people IIRC have actually reverse engineered that
> bit of the firmware, but it has not been incorporated yet into
> the Linux version of the drivers.

Indeed - mostly because of the attitude of the madwifi authors, AFAICT.

> mike> Also, in my experience, they don't work at all in ad-hoc
> mike> mode, which is a pretty major bug IMO. [ ... ]
> 
> Well, it does not look that major to me -- most recent drivers
> barely work, never mind in ad-hoc mode (or AP mode) :-).

Hardly a good thing. I'd really like someone to explain to me how
OpenBSD can have such good free support for wireless hardware, when the
Linux situation is a nightmare of often non-free, different drivers with
varying levels of support that are buggy as hell to boot. And Linux
developers (as we're often told) are supposedly more concerned with
free-as-in-freedom code than BSD people. If Debian GNU/OpenBSD ever
makes an appearance I'll be sorely tempted. Theo DeRaadt does more to
wring specs out of recalcitrant hardware vendors than anyone I can think
of on the Linux side of things, and he actually gets results.

Mike.

-- 
Gllug mailing list  -  Gllug at gllug.org.uk
http://lists.gllug.org.uk/mailman/listinfo/gllug




More information about the GLLUG mailing list