Ubuntu on a 486?

[Adam Conrad]

Adriano Varoli Piazza wrote:
> 
> Adding to my other mail, there's an important difference between
> these machines and 486 hardware: you can find them on the market,
> new. They aren't available only from thrift shops, garage sales, they
> come brand-new from Sun. Or is there a shop where you can get new 486
> boards, processors, etc?

Yes, it's been pointed out several times in this thread that there are a
large number of 486-class machines still being sold.  The AMD "Elan"
CPUs are 486-class, and found in a large number of devices on the
market.  I own several fairly new 486-class machines that I use as
routers, Oliver spoke of new 486-class thin client boxes, etc.

> Well, the obvious answer would be "the developers Canonical pays for 
> this would then be able to do something else to provide us with the 
> best distro on the desktop", or, failing that (it's better to pay 
> someone to do a job he/she already likes) "Canonical could pay 
> developers to focus more on their specifical objectives".

I think you make the mistake here of believing that we expend a great
deal of energy targetting 486-class CPUs.  If a bug comes up in the
toolchain where we find we're accidentally generating opcodes a 486
can't handle, sure, we'll look into it, but for the most part, the work
here is twofold:

1) Build a kernel that can run on 486 machines (the poorly-named -386
kernel, in our case... This is reasonably trivial)
2) Make sure our toolchain is outputting code for 486 CPUs and above,
which is a matter of setting the default arch and tune flags, and we're
done.  Again, trivial.

So, the few times we'd have to kill opcode bugs, sure, that time could
have been spent elsewhere, but I suspect it's nowhere near as much time
as you think it is and, frankly, our toolchain hackers probably just
want to hack on the toolchain anyway, not fix bugs in evince. :)

... Adam