Greg K-H Recommends New Kernel Version Naming

Oct 17, 2008

Greg Kroah-Hartman, Linux developer at Novell, suggests a new naming scheme for the Kernel releases on the Kernel mailing list. The four-digit year would be included in the name.

Kroah-Hartman writes that he had meant to bring up his suggestion at the Kernel Summit 2008 in September that he does now as a follow-up. He had always been involved with version numbering and finds the current scheme unmanageable.

His recommendation: future versions of the Linux kernel should follow the pattern four_digit_year.release_number.minor_release_number. By this pattern the first release in the coming year would be 2009.0.0, the second 2009.1.0, and so forth. If zero-numbered releases are undesirable, he says, the year could also begin with release 2009.1.1.

Kroah-Hartman points out that the naming scheme provides a better way to determine how old the release is. Kernel 2004.9.0 would be easier to date than Kernel 2.6.9.

The current scheme follows the pattern 2. major_release.minor_release.extra_version. With the new scheme, kernel macros with major and minor numbers would also have compatible naming so as not to break scripts, he assures the community in his email.

Among the first responses was one from H. Peter Anvin who found it easiest just to increment the last digit, such as 27, 28, 29, etc. He finds the existing prefix 2 with the large following digits to have "outlived its utility" and suggests bumping the version to 3.0 and maintaining that scheme for "huge changes."

Adrian Bunk had rather a different view and warned against radical changes that could affect not only the Kernel and its tools, but break countless userspace programs. Many of these programs interpret the kernel version number through calls such as uname -r. He gives a snippet from an OpenSSL library script as an example.

Related content

Comments

  • Age of release

    "Kroah-Hartman points out that the naming scheme provides a better way to determine how old the release is. Kernel 2004.9.0 would be easier to date than Kernel 2.6.9."

    Even better would be YYYY.MM.1.
  • nice idea but,

    There is still a lot of stuff at kernel eco-system that can make a dramatic difference in user experience. udev and HAL being two examples of making solid improvements over the programs they replaced. Xorg is another program that can do with a lot of help. Unfortunately, only Redhat, Novell and IBM are contributing significantly to the plumbing that benefits everyone. Linux kernel itself attracts a lot of attention but the packages around it don't match the pace.
  • most sensible out of all

    To date, this is most sensible proposal.

    Though I'm one of those who really thinks that changing now kernel versioning provides little to no gains.

    To me it more and more sounds like lots of people are bothered by visibly smaller role kernel now is playing in the whole "Linux OS" thing. Seems like people want some shake up of kernel development to make their contributions look more significant.
  • I see the flame war starting...

    Don't think the new proposal will pass however.
comments powered by Disqus