Is any attempt to standardize Linux akin to herding cats, especially given the proliferation of distributions and packages? Jim Zemlin doesn't think so, and has been trying to do something about it via the Linux Standards Base.

I talked with Jim Zemlin about the LSB back at the Red Hat Summit, and according to his description of it, the incarnation they had at the time was mainly intended to insure that server applications had a stable platform to build on. Desktop apps weren't part of the picture, but after reading a piece in InternetNews about the LSB, I realize that was just a question of scope for that iteration of the LSB.

The newest version of the LSB, set to be finalized by November of this year or so, is about stabilizing the APIs that are used in the major distributions (Red Hat, SUSE, Ubuntu), and so on. Many of the things in the version 4 road map have been badly in need of some kind of overall plan: high-level frameworks for multimedia, for instance. (ALSA, GStreamer and PulseAudio are specifically mentioned in the 4.0 map.) This part I'm especially curious about, since it represents one of the areas where Linux still has a lot of progress to make, apart from the efforts of individual projects.

One key component of the LSB is a test suite that anyone can use to determine if a given application they've put together is LSB-compliant. There's no way to guarantee that someone else will hew to the LSB, but Jim's idea is that anyone who cares about having their application run seamlessly on most editions of Linux, will take the effort to run the test suite. And by making the bar for compliance as low as possible, there'll be little reason not to. In short, not complying with the LSB will be its own punishment.

Is that the most effective way to do it? In the Linux world, it might well be -- but the real proof will be to see how much adherence there is to the 4.0 standard a year or more after it's been published. I'm also wondering whether the more maverick distributions will even think of this as something to hew to, or to dissent from. After all, if Linux's big defining trait is its diversity, will that include diverging from things that are meant to give the community that much more solidarity?