On Fri, Sep 01, 2006 at 12:16:59PM -0700, Spruell, Darren-Perot wrote: > From: Charles M. Hannum [mailto:[EMAIL PROTECTED] > > On Fri, Sep 01, 2006 at 10:40:01AM -0700, Spruell, Darren-Perot wrote: > > > Like, what docs does a vendor engineering division give to the > > > developers who write the drivers internally? They don't > > give them bad > > > docs. They give them functional, useful docs. Does it need to be > > > stated that any project wanting to compose useful support > > for the same > > > hardware shouldn't get the same level of docs? > > > > Sorry, but that's the core fallacy in your argument. In many > > cases, there are no "functional, useful docs". They just > > don't exist. Certainly this is a problem in itself. > > Certainly it is. So why bother resorting to vendor-supplied drivers (OSS or > blob) derived from originally piss-poor docs in the first place? If the docs > are bad, then the results of those docs are derivatively worse as a result.
That's not actually true. You're still using the fallacy that the vendor driver is written based on the documentation -- but in fact there are other inputs, like discussion with the hardware engineers. Sometimes there are pieces you just can't get from the documentation, because they're not there, but they are present in the driver. In the current climate, having both is almost always better than having only one -- and certainly having the code is better than having nothing. I'm not against harassing the hardware vendors to do better.
