See also: last week's Kernel page.

Kernel development

Pre-patch 12 for the 2.0.36 stable kernel release is out, see the announcement for details. This one has also been sent off to Linus for consideration as the real 2.0.36 release; no answer so far.

Why did 2.1.123 come out with compilation errors in the frame buffer console code? The discussion on this topic has been heated, acrimonious, and without resolution so far. The basic cause of the problem was that Linus applied some patches to the include directories while leaving out a substantial number of changes to the video drivers. As a result the C source and header files do not match and things do not build.

There are disagreements as to how the patches got dropped, whose fault it is, etc. Strong disagreements. The discussion reached a point where Linus has thrown up his hands and stated his intention to take a vacation for a few days. He appears fried, many of the important kernel developers are frustrated, people are mad at each other, and real work seems to have stopped for the time being. It is not a good situation.

The core of the problem seems to be that Linus is just getting overwhelmed by the volume of patches that he is receiving. The kernel is getting larger and more complex, and, as they say, "Linus does not scale." As a result he drops a lot of incoming patches, leaving it up to the developers to monitor the kernel releases to see if anything has happened yet. Since Linus is unwilling to delegate the task of reviewing and applying patches, the problem looks difficult to solve.

Some thought is going into a solution; with luck, something will be found that is acceptable to Linus. Hopefully that, and the eventual release of the 2.2 kernel (and subsequent return to fun development work) will help to ease tensions before something serious gives way.

Discussion continues on the Linux scheduler and whether a separate queue should be created for real-time tasks. Richard Gooch has run a series of scheduler benchmarks; the results and his interpretation of them can be found in his web site. Richard also posted a patch to 2.1.123 which contains a number of scheduler tweaks, and implements the real time run queue.

Work still continues on the NFS server in 2.1. H. J. Lu, Bill Hawes, and G. Allen Morris are all contributing fixes. Nonetheless, the kernel NFS implementation still lacks the stability it needs, has interoperability problems, and NFS over TCP does not work well at all. H. J. Lu has a patch out with the latest fixes. People interested in NFS should really consider trying it out.

It is important that 2.2 have a good NFS implementation. It's bad enough that NFS v3 will not be supported; at least the v2 support should work well. Unfortunately, there is currently nobody who is really taking responsibility for Linux NFS.

Just one thing for those who have not seen enough UDI talk yet: Kevin Quick, leader of the UDI project, posted a simple explanation of UDI for those who don't want to dig through the UDI web site. He does this by comparing UDI to the VHS video standard. The analogy kind of works for a while, but then he heads into I2O stuff: "...let's say that I2O is a new audio encoding format designed to provide surround sound separate channels and superlative quality and signal-to-noise ratios." Whatever...

Since we're a weekly publication, chances are we'll be behind a rev or two on the kernel release by the time you read this page. Up-to-the-second information can always be found at LinuxHQ.

Next: Distributions