[PLUG] Fedora or Red Hat (or OSX)

[Aaron Burt]

[Warning: long-winded answer.]
On Thu, Dec 11, 2003 at 09:16:20PM -0800, Carla Schroder wrote:
> And don't forget the reason for poor hardware support in Linux is retard 
> hardware manufacturers, who I suspect are so deeply in Microsoft's 
> pocket they dare not support Linux. Otherwise how can anyone possibly 
> explain their continued refusal to make use of the huge free Linux 
> talent pool, which toils to write drivers even in the face of 
> opposition? They don't even have to write their own, it's already being 
> done. But too many of them won't even supply specs, or product for 
> testing. 

Three major reasons:
1. IP entanglement.  Most devices are made with chips from a few
different sources, and info about how they work is tied up in NDAs and
other contractural agreements.  It takes money, time (money), lawyers
(money) and internal politicking (money) to untangle things enough to
allow for open-source drivers and/or open specs.
 
2. Communication issues.  Oftentimes, components or even whole devices
are bought via a middleman or from a manufacturer who may feel that they
put their money into supplying a set of generic drivers and manual just
so they wouldn't have to deal with OEMs bugging their engineers.  The
engineers may be perfectly willing to supply the info to someone who's
capable of doing the drivers, but technical companies are organized to
keep the pesky customers from bugging the engineers.

3. Marketing issues.  Many important decisions end up being made in
Marketing, including things like platform support and target market.  
It may seem weird, but I hear that in some companies, when Marketing
says "XP and Win2K only", they mean it, and don't like folks messing
with their decisions.

So, yeah, stupid.

Oh, and don't forget the whole issue of hacking.  I'm sure manufacturers
worry that badly-written drivers could break hardware and cause warranty
claims, and I hear Intel doesn't like the fact that open-source drivers
for their Centrino WiFi chipset could allow for illegal transmit power
levels.

It's changing, though.  And I think that as long as it's easy to supply
specs or GPL code to someone who'll get it into the kernel, and really
heard to ship a binary driver for Linux, it'll continue to change.

Still, I wish there were a li'l library of reverse-engineering tools
(various sorts of bus scanners, etc.) to help with driver development.
Heck, maybe some well-funded Linux-booster organization could fund a
couple driver hackers and give 'em workbenches...