Actually, IMVHO, people still should run custom kernels because otherwise they turn into sissies who waste hours switching distributions every time they hit a driver bug.
I think a strong case for "regular" Linux users being able to build there own kernels is that while most users don't need or want a "bleeding edge" userland, they often want a bleeding edge kernel. For example, I use Debian Stable because I hate having inconsequential stuff change out from under me every 6 months, but I need up-to-date kernels for my hardware to work.
That is a niche (a large one I think) that is unfilled by the popular distros as far as I can tell.
Seven months later, Debian 3.0 was released - with Python 2.1.
...time passes...
In November 2004, Python 2.4 is released. Debian stable is still using Python 2.1, and will continue with it up to June 2005.
June 2005, Debian stable finally upgrades off its 4+ year old Python version.
Needless to say, many 2.2+ scripts I ran in early 2005 broke on my Debian stable.
I understand that the late date of that particular Debian release was partially due to Ubuntu hiring away developers. I also understand that I could have always ran unstable. But it put me off of Debian stable.
Hmm, I can see that being a problem. Though I have a habit of avoiding distro-packaged interpreters/compilers anyway, even with Fedora, so that I'm always on the latest.
