Tuesday, March 31, 2009

Moving forward

At the end of my involvement in FreeDOS, I ask myself: What does FreeDOS need to move forward? What should FreeDOS do to make it stand out?

First, I think the FreeDOS kernel needs to be cleaned up. There has been a lot of tinkering in the kernel, trying to save 10 bytes here, 8 bytes there, 12 bytes somewhere else. While the motive was good (kernel should take up the least memory possible) what's happened is that the kernel source code became more difficult to work with. DOS should be a fairly simple thing, compared to other kernels. I want to encourage some interested developer to pick up the source code and simplify it, even if that means undoing any of the memory savings.

Above all, the kernel needs to work reliably. Today, we have two branches of the FreeDOS kernel: 2036 stable, and 2037 devel (unstable). That shouldn't be ok, yet somehow we've convinced ourselves this is acceptable. Having two versions of the kernel, where the most recent branch is effectively "broken", is what's keeping us from moving forward with the kernel. Is it easier for a kernel developer to start with 2036 and re-add the features from 2037? Or is it better to fix the broken parts from 2037, to release a (working) 2038 version? I lack the skill to do any kernel development, so I never tried. I'm hoping someone with the necessary energy and enthusiasm can work it out.

No comments:

Post a Comment