Accessing the BK2CVS repository
People do use the CVS repository, however. It just turns out that many of them have noted that it is faster to use rsync to update the entire repository from a kernel.org mirror than to update it through CVS. The rsync approach looks like the way to go in the future, but it does have one potential difficulty: if the repository is updated in the middle of an rsync, the person downloading the copy might get an inconsistent tree. Kernel hackers have to deal with enough race conditions as it is; they would prefer not to encounter them while trying to update their copy of the mainline kernel repository.
The solution that is likely to be
implemented involves the creation of a couple of sequence files. One is
fetched before doing the big repository rsync, and the other afterward. If
the sequence numbers in the two files do not match, the rsync operation
raced with an update of the repository and needs to be retried. This is,
of course, an Internet implementation of the seqlock algorithm used within the kernel. Look
for an update script to show up soon.
