[mythtv-users] OPen source video card support (Was:Is NVIDIA worth the bother?)

[Dale Pontius]

Marc Randolph wrote:
> On Fri, Sep 25, 2009 at 2:14 PM, Johnny <jarpublic at gmail.com> wrote:
>>> I don't know how much HDCP is a factor in keeping the API closed -
>>> after all, Intel is doing their best to open their video stuff up.  Of
>>> course, even if HDCP didn't exist at all, competitive pressures would
>>> likely keep Nvidia and AMD/ATI stuff closed - for the same reasons
>>> that Broadcom and Marvell keep their stuff closed.  If they think that
>>> exposing any information (be it APIs, or datasheets, or whatever)
>>> would help the competition in even the slightest way, their deluded
>>> paranoia usually prevents them from opening up at all.
>> ATI has opened most of their specs as of a year or two ago. The open
>> source community just hasn't been able to produce a quality driver on
>> the same pace as nvidia's closed source drivers. I was at an event
>> where an ATI engineer was speaking about their hardware and Richard
>> Stallman was protesting. The engineer said that the main issue with
>> opening the spec is that unlike the CPU community, they make radical
>> hardware and API changes from one GPU to the next. At the time they
>> simply don't want to deal with the legacy support and complexity that
>> came with opening it up. They felt it would make it harder for them to
>> make radical changes as quickly as the currently do. Second simply
>> producing and documenting all the APIs etc and making them available
>> with the appropriate licenses is time consuming and they simply don't
>> have a high incentive to do the work.
> 
> Very good point.  I've experienced first hand that an additional
> reason that companies keep things "closed" is so that they don't have
> to support them.  The small company I worked for previously was
> interested in using one of Broadcom's larger next-gen switch chips,
> but it was explained to us that since they had just a small support
> group for that product, they only allowed a few customers to have
> access to it (namely the big boys... Dell, Cisco, Juniper, etc).
> 
Now that you mention it, I've not just been there, I AM there.  Not in
the GPU or CPU business, but still in the chip design business.  We get
a design done, debugged, qualified, and then we're on to the next
technology node.  We do end up "going back" from time to time to take
care of issues, but as a gleaming generality, we don't like to.  It's a
distraction from development of the next generation.  We develop specs
and documentation for both external and internal (primarily
characterization) use, but there's a world of difference between the
two.  In general, external documents have to be written to a higher
standard and go past more eyeballs on their way to customers, including
lawyers' eyeballs.

External release of thorough internal documentation is not a minor
effort.  The REAL crime here is that it's not "normal practice", as it
is for CPUs, for instance.  These days, "documentation" is the pdf file
telling how to install the Windows drivers, and "support" is the drivers
themselves.  That's the crux of the problem.

Dale