Part 2 Of Nouveau Saga: The Microcode

[yoshi314]

His point is valid, the OSS community tends to improve faster when the code is consolidated upstream rather than having everyone maintain their individual patch sets.

just a few months ago i would nitpick that linus wants as much as possible developed in-tree, but he rejects crap code.

fortunately staging drivers feature appeared and it makes it possible to develop incomplete/testing code in-kernel, so i find his point valid.

[nibor]

"I will not link to SourceForge mailing list archives."

They are crap to navigate. In fact, I have no idea how I can easily go to previous/next message in the thread there.

The markmail.org interface is really nice. Here's a link to the nouveau discussion:

http://markmail.org/message/wgczkuh55ehvcwsw

Their search and filtering is also pretty good.

[RahulSundaram]

The part that makes no sense to me is that apparently Red Hat's lawyers are saying it's fine to develop this ourselves inside Fedora, but not in the official Linux kernel which we don't control. I mean, in any reasonable universe, wouldn't it be the opposite? I would think that putting it in Fedora only would mean more legal risk to them, and it would be less if it was in someone elses repository.

You have certainly missed a key part of the debate. Dave Airlie from Red Hat explained this at

http://sourceforge.net/mailarchive/m...mail.gmail.com

[Xheyther]

It look like the nouveau driver was finally pulled into staging (at least linus was finally requested to do so : http://sourceforge.net/mailarchive/m...ynet.skynet.ie).

[ernodv]

It look like the nouveau driver was finally pulled into staging (at least linus was finally requested to do so : http://sourceforge.net/mailarchive/m...ynet.skynet.ie).

Nice. If it goes in, maybe Ubuntu will re-think their decision with 10.04LTS&2.6.32.

[Craig73]

It's interesting how quickly, one the roadblocks are identified, something can be resolved (at least in part)

[DeepDayze]

Reverse engineering does not protect you as far as possible patent restrictions. If nvidia for example has a patent on a process and that the reversed engineered firmware duplicates obtained in a clean room environment, from scratch or copying knowingly or not, it may very well infringe on patents made by nvidia or partners they license from. A simplified example:

Two guys are inventors, they both are developing mouse traps. Guy A has a piece of wood, a spring, and some stiff wire and comes up with the spring trap and patents it. Guy B with the same materials without even knowing it has already been patented comes up with a design looking very similar in looks and functionality and uses the same principals. Guy B is in patent infringement if he tries to market it (price really isn't a issue here, could be a freebie or charging for it).

Bottom line is patent infringement, no matter if it is ignorant of the patent or not, is still illegal in the eyes of patent law.

Good point. Just what I meant in my earlier post. Nvidia should say something about this issue as well. If there's IP in the microcode they want to protect, whether theirs or the 3rd parties, they need to say something so the developers are aware of the patent implications

I would suggest that the firmware be at least nonfree and restricted...and kept in a separate package or at least be made available for download from appropriate locations

[deanjo]

Good point. Just what I meant in my earlier post. Nvidia should say something about this issue as well. If there's IP in the microcode they want to protect, whether theirs or the 3rd parties, they need to say something so the developers are aware of the patent implications

True, nvidia are the only ones that can verify this. It would be in the best interest of the nouveau devs to be pro-active and ask nvidia to take a look at it to check for these possible infringements. If none are found then fine. If there is then see if some sort of solution like On2 and Xiph.org did with the patents utilized in theora.

[DeepDayze]

True, nvidia are the only ones that can verify this. It would be in the best interest of the nouveau devs to be pro-active and ask nvidia to take a look at it to check for these possible infringements. If none are found then fine. If there is then see if some sort of solution like On2 and Xiph.org did with the patents utilized in theora.

That'll be the good way to go about it. If nvidia gives its blessing then all's good

[feefref09]

Ubuntu vote here. Ive tried RedHat and Fedora in the past but just didnt get on as well them.

I like ubuntus wiki as well which has probably helped my learning process.