Re: Kernel-WhyTo
On Sun, 16 Aug 1998, Matthew Myers wrote:
> I am curious about why there are 9 Million different kernels. I was
> examining the 2.1 kernels and they number all the way up to .115. What is
> the point of this? If .95 is obviously an improved and better version than
> .94, why are people still trying to work bugs our of .95? Why not go on to
> the next and work on it?
First off, .95 is *not* obviously an improved version compared to .94.
Often completely broken code is merged into the development kernels, stuff
that won't even compile, just so that others can review progress and make
suggestions.
Second, sometimes you need a stable base to work from. You can't always
be sure that a bug is the result of the changes you made or a bad effect
from someone else's change or both. If you stick with one version for a
while, you can get your code straightened out and mostly debugged, then
worry about matching up the interfaces to other people's code later.
Of course, theoretically the *2.0.x* series gets better as x increases,
because that's the 'stable' branch and only bugfixes are supposed to be
merged in. As every developer learns (sooner rather than later) fixing one
bug can introduce another. Some paranoid sysadmins (often paranoid because
they are paid to be paranoid) find a kernel version that works, and stick
with it. They know the bugs in, say, 2.0.30, but none of them apply to
their installation and they're scared of what unknown bugs might lurk in
2.0.35.
Sincerely,
Ray Ingles (248) 377-7735 ray.ingles@fanucrobotics.com
"The meek can *have* the Earth. The rest of us are going to the stars!"
- Robert A. Heinlein
Reply to: