Devel Digest, Vol 38, Issue 1

[Mitch Bradley]

Martin Langhoff wrote:
> On Wed, Apr 1, 2009 at 6:26 PM, Mitch Bradley <wmb at laptop.org> wrote:
>   
>>> that's the great thing about linux -- just when we finally
>>> develop a standard way of doing things, someone steps in and
>>> says, "no, that was wrong, let's do it this way instead", and
>>> stops supporting or improving the old way.
>>>       
>> ... and then people wonder why so many hardware companies don't want to
>> support Linux with drivers.
>>     
>
> That's not entirely fair. If you submit the driver for inclusion in
> the kernel tree, then the dev who pushes for the API change is on the
> hook to update all the drivers to the new API.
>   
Believe me, I'm well acquainted with that argument.  But submitting the 
driver to the kernel tree requires a fundamental change in mindset for 
most companies in the competitive commercial landscape. They have 
trouble with the fact that there are other companies lurking around the 
corner wanting to take away their business by cloning their designs.

Furthermore, the explicit disclaimer of warranty that underlies much of 
the FOSS legal underpinnings is counter to the fundamental business 
realities in the commercial world. Pushing a driver upstream and saying 
"upstream will then take care of it for me" doesn't "solve the problem" 
when you consider that companies have customers that expect a complete 
solution and are always working on new hardware versions.  If they are 
going to "support Linux" for real, it's necessary to port the driver to 
the new version, or wait for upstream to port it.  The former is 
problematic if the upstream churn has invalidated knowledge that your 
driver expert used to have, the later is problematic from the standpoint 
of managing your internal schedules and keeping your plans secret from 
your competitors.

It's nice to say they should "see the light", but in my experience 
talking to many such companies, the fact of the matter is that it is a 
hard nut for them to swallow.

I see their point.  I have been doing Unix kernel work off and on since 
1979.  I used to enjoy it; now I dread it.  Unless I do nothing else, my 
knowledge of the necessary details becomes obsolete so fast that every 
project makes me feel like a n00b.