m68k support

I guess it's time for a rant.

One question that keeps popping up regarding m68k support is "why do you guys compile package X? No sane person is going to run that on m68k anyway!"

I've had to refute that point on IRC saturday, and in private mail just today.

Well, maybe you're right. Maybe nobody is interested in running mozilla, KDE, GNOME, OpenOffice.org, or theorem solvers on m68k. (I hadn't even heard about theorem solvers before saturday; apparently, a theorem solver is something that tries to prove a theorem by starting with some axioms and some rules, and may take days/weeks/months/years to finish on even the fastest computers.)

But the question we need to ask is not "Would I, personally, have any use for this software package on m68k"; rather, the right question is "Would any single user in our user base be interested in having this software package?" That is the only right question, because that's what puts our users in the center of the argument; and according to the Social Contract, section 4, our users are one of our two priorities.

The problem with that question, however, is that we cannot reasonably answer that question. We cannot know who our users are or will be, today and in the future. Nor can we know what our users use their system for. Some might use it as a home server, and might require server software. Some might want to use it to show their friends how cool this GNU/Linux thing is, that it even supports the most recent software on outdated hardware. Others might have other reasons to run GNU/Linux on their m68k box.

Therefore, since we cannot reasonably answer that question, we have only two options: either we compile all applications on all architectures, including m68k, where technically feasible and possible (compiling lilo on m68k is not "technically feasible", even if it's technically possible; and in the case of OpenOffice.org, it's not technically possible (OpenOffice.org requires architecture-specific assembly code. Yes, I tried :)), including those that seem silly, or we compile none of them and discontinue the m68k port entirely.

Some that read that last option might be surprised to read something like that coming from me (Hi Md :). Yes, I do agree that we will probably not be able to support the m68k port "until the end of times". But let's make one thing clear: as long as we can keep up building packages for m68k and continue to have users (even today, there are still people who install Linux/m68k for the first time), we should continue to support the m68k port.