New nvidia driver obstacle

Jonas Sundström jonas at
Wed Apr 13 09:19:44 PDT 2005

Miguel Mendez <flynn at xxxxxxxxxxxxxxxxxx> wrote:
> It's interesting but it looks like he's basing his code on the old 
> GLX code, so I'm not sure if he'll be able to support anything newer
> than the GeForce1 gpu.

I think he explains this in the entry of March 3'rd. I believe
he wanted to start from a combination of parts known to 
work together, on BeOS, and take small steps forward. 

The posix-but-not-quite-unix platforms are clearly at a 
disadvantage, in that we can't just recompile X & Co.
and be done. The latest everything compiles on Linux. 
But we're not Linux. :I

> It would be great if nVidia could release the specs 
> for their old cards, but given their track record it's quite
> unlikely. You can always get an ATI card :-)

Are ATI any better at releasing spec?

> I used to run BeOS 4.5 as my main desktop for a while 
> and there are some ideas from BeOS that might be 
> adapted to DragonFly. I'm thinking rich metadata mainly 
> and some parts of their API.

Yeah, the filesystem goodies of BeOS just make sense,
at least for desktop use. I believe POSIX extended attributes
could be used similarly, except they're untyped, (I think), which 
complicates things, and there's no live indexing facility. Perhaps 
the journaling mechanism will allow for wiretapping filesystem 
metadata, to allow for stuff similar to BeOS's (live-) queries, 
MacOS X' Spotlight, Linux' Beagle and whatever Microsoft 
come up with in place of WinFS.

/Jonas Sundström.        

More information about the Kernel mailing list