Replies

The guy who had added the patch was Greg Kroah-Hartman (Greg KH) who is the maintainer for the tree holding many of the drivers - pieces of the Linux kernel that deal with specific hardware, such as video, sound and network cards.

Andrew Morton had agreed, resulting in the "I think I'd favor that" quote opening this posting. Andrew is number two in the Linux world these days, managing the largest of the merge trees that feeds to Linus.

Then Linus Torvalds came out strongly against the patch, as noted in the second post above. Linus is still top dog in the Linux world, integrating the main feeder trees from Greg, Andrew and others into the main Linux tree, which is then used as the basis for many distributions such as Red Hat, Debian, SUSE, Ubuntu, and a cast of hundreds. Linus announced that he would refuse to accept this patch unless it was -first- accepted into these major distributions (a reversal of the usual order of patch traversal.)

Others chimed in, several agreeing with Linus.

Greg KH was persuaded that his patch was wrong, though his frustrations with companies abusing the Linux GPL license remain. He pulled his patch from his driver tree, resulting in Post 4, above.

This patch is clearly dead.

Nvidia currently keeps the bulk of its kernel driver code for Nvidia video cards closed source and proprietary. It loads a driver stub into the kernel that is open source but not GPL. You can get the source for this stub, but it is not GPL licensed. That stub is small, and has no particular knowledge of Nvidia video hardware. That stub loads as a driver into the kernel, and interfaces with a closed source proprietary module that has Nvidia secrets and knows how to driver their video hardware.

ATI (now owned by AMD) and Intel (who actually produce the largest quantity of video graphics chips) do similar things to keep their video hardware secret.

It is not clear how these many hardware vendors, working with closed source drivers and trade secrets critical to their competitive success, would have responded to the GPL-only driver patch, had it been taken into the main kernel tree and major distributions. Perhaps much of the most interesting video, networking and storage hardware would have become unavailable to the Linux market, which would have been a big problem for those of us working in or depending on that market.

Correction - well, clarification:

main Linux tree, which is then used as the basis for many distributions

should have been:

main Linux tree, which is then used as the basis for the kernel in many distributions

Certainly the bulk of any major distribution is not the kernel, but rather the many other pieces of software such as the windowing (X11, KDE, Gnome, ...), desktop, application, build (compiler, make, ...) and many other packages.

Excellent analysis of this issue. I think it's great how the development effort behind Linux is completely open to outside scrutiny. Arguments for or against proposals are made, and some times, consensus reached without totally blowing things up. I've seen discussions about other issues that resulted in major project splits that forked off for a while, because Linus wasn't convinced of the utility or stability of a given approach that later were successfully merged back into the official tree after the ideas had proven their worth and utility.

To me, this freaking rocks, as we, the end users, end up with the best of both worlds. You can follow a fork and use it if you want, or you can use the main supported kernel if you feel more comfortable with it. No one tells you that you must do things their way. Well, they can tell you all they want, but you don't have to listen :-)