This is a nice theory. My worry is that this may not be like that
in practice. Samba is a good counterexample (not sbmfs only).
It is generally regarded as a pretty portable code. Red Hat had to
issue errata as it broke in 5.0 distribution on Alpha. If a code
works on practically everything but only on **some** Linux systems
then I am afraid it will be hard to convince developers that it should
be fixed; even if this "some" will turn gradually into "most".
Here you have some examples which at least will have to be checked
carefuly: Perl (its code is pretty clean in general, so maybe it will
not cause problems), Tcl (all official distribution I know so far are
broken for other reasons, although bugs are masked on 32-bit machines,
and I do not see a big rush to fix them), Apache, (g)awk... This
list can be continued.
> Yes, you've seen breakage reported on this list - but this is a development
> list, and breakage of one sort or another gets reported here pretty often.
If I would see these reports only on this list I would not worry so much
(although some problems seem to have an amazing staying power). But
I see related questions on axp-list and they probably start to spread
as Linux/i386 will begin to use glibc seriously.
> It buys us fewer headaches and greater flexibility in the future, at the
> admitted expense of some headaches now.
I would really like that this would be so. So far the evidence
which I have seen was not very convincing. And once again, I am
not talking about linux-kernel list only. I know that this is
the place to report troubles. :-)
Michal