[gta02-core] OT: binary drivers

Álvaro Lopes alvieboy at alvie.com
Sat Aug 29 15:35:50 CEST 2009

Dave Ball wrote:
> Christopher Friedt wrote:
>> On Wed, Aug 26, 2009 at 11:20 PM, Dave Ball <openmoko at underhand.org 
>> <mailto:openmoko at underhand.org>> wrote:
>>     Ack. Binary is evil
>> I agree - binary drivers are a evil (but still better than nothing for 
>> the average end-user who never sees a line of source code).

Binary drivers (actually mixed drivers) are often the only way manufacturers have to protect their design effort. This is very unfortunate, but it is somehow
understandable. Competition is fierce. IP/Patents are often not enough to protect the investment made in the product. Like:

 * you spend 12 months designing a good chip, invest in 8-10 people, you'll invest about 300K. You sell each device $25 a piece.
 * you release open drivers, others will figure out your design internals. They work hard (and cheap).
 * Some chinese company offer 100% HW compatibility with your design, each piece costs $2. This in just 3 or 4 months.
 * You realise it was not worth the effort.

I am not standing for closed-source, binary drivers, just stating I can understand their point (I cannot understand Calypso, it's an outdated chip, why closing
its specs ? - or maybe it is still selling like bananas).

> In an embeded system such as ours, and a community project such as ours, 
> it seems to me that the problems with binary drivers are particularly 
> apparent.
> We'll need to make changes to our kernel to support the new board, and 
> debug kernel problems we come across.  Dealing with a binary module or a 
> fixed ABI version would (imho) make things unnecessarily painful.

True. And would defy the basic principle - openness.

> There might be the possibility of one or two individuals getting NDA'd 
> access to docs and/or an existing binary driver, and can tweak or 
> up-level it for us - although not having a big corporation and pockets 
> full of cash makes this tricky too.
> At the end of the day, some people do decide to use binary drivers - but 
> my opinion is that is counter to the 'open' aims of our project, and 
> we've got better things to focus on.  To me, a binary driver would be 
> the same as no driver - we couldn't support it, and trying to support 
> any future HW through a binary driver is likely to be a distraction from 
> the rest of the system.

Unless we get proper API documentation for that driver. I actually do not care much if my nVidia drivers are binary and closed. Their OpenGL API is enough for
me to get the best from the card.

> Still, we've got a long way to go with our current HW before considering 
> new devices, so no need to rule out any chips yet :-)

Now that's talking :)


More information about the gta02-core mailing list