What's a good video card?

[Aleksandar Milivojevic]

Rick Stevens wrote:
> Then stop upgrading the kernel.  M$ doesn't update their kernel as often
> as Linux does so ancient Windows drivers on old hardware stick around
> longer.  I've even heard people buy new hardware, then whine because
> the maker doesn't provide a driver for Winblows 3.11!  IMHO, Windows
> drivers have an artificially long lifespan due to M$ not fixing things
> that are known to be broken.  Linux is more fluid because the kernel
> gang never stops tweaking things in an effort to make things better,
> faster, more reliable, etc., etc.

Rick, you are mixing apples, bananas and oranges here.  Mixing them 
might only produce interesting juice drink or fruit sallad.  Nothing more.

Lifespan of Windows operating systems (all 2 1/2 of them: DOS, Win9x (I 
count this one as 1/2), and NT) has nothing to do with Microsoft fixing 
or not fixing the bugs, or longevity (or lack of) device driver API. 
You can fix the bugs and have stable device driver API.  You can never 
fix bugs and have unstable device driver API.  One has nothing to do 
with another.  As an example of operating system where bugs are fixed, 
new features and improvements added regullary, and stable device driver 
API is maintained, take a look at Solaris.  Commercial companies don't 
like unstable APIs.  And that is the real (and only) reason why Linux is 
so poorly supported by them.  I find it sad that Linux developers have 
choosen hard-line approach in regards to companies that actually make 
hardware.  Hoping they will embrace open-source as the only way to make 
software is silly (to say it politely).  Ultimatums never yield 
solutions (or at least not good/optimal solutions).  Especially not when 
you don't have a big stick to wield around when making ultimatums 
(remember, Microsoft is the one holding that big stick right now and for 
forseable future).

And yes, I strongly beleive that unstable device driver API in Linux is 
more result of political (as opposed to technical) decisions.  There's 
really not that many ways interaction between kernel core and device 
driver can be implemented to handle an interrupt, or schedule DMA 
transfer, or access device's registers and/or memory.  And most of them 
are really not any better than the others.

-- 
Aleksandar Milivojevic <amilivojevic at pbl.ca>    Pollard Banknote Limited
Systems Administrator                           1499 Buffalo Place
Tel: (204) 474-2323 ext 276                     Winnipeg, MB  R3T 1L7