Re: Kernel interface changes (was Re: cdrecord problems on

Oscar Levi (elf@buici.com)
Tue, 9 Feb 1999 14:16:13 -0800


On Mon, Feb 08, 1999 at 12:32:52PM -0800, david parsons wrote:
> In article <linux.kernel.19990208112914.A7925@hazel.buici.com>,
> Oscar Levi <elf@buici.com> wrote:
> >On Sun, Feb 07, 1999 at 08:15:24AM +0000, Linus Torvalds wrote:
> >> I _refuse_ to even consider tying my hands over some binary-only module.
> >
> >Hear hear. Deviating from binary-only compatible is a significant
> >part of what makes the Linux kernel work. "If it isn't broken, don't
> >fix it. It it is broken, do."
>
> That's a nice sentiment, but if you break binary compatability
> (which is useful for modules you have the source for as well; a
> fixed binary interface (like, umm, UDI) makes it easier to mix and
> match device drivers and not be forced to follow kernel revisions or
> do lots of hackery if you want to support something, like the xia
> filesystem, that's not supported in the kernel de jour), you're
> only fixing it if you consider backwards compatability to be a bug.

I don't believe that mixing drivers from different versions is
important enough to retain any given interface. Backward
compatibility is not an absolute benefit. As we learn to improve
interfaces, the flexibility to change them is paramount. In
user-mode, where compatibility is judged to be *very* important, the
kernel entry points have maintained long-term backward compatibility.

UDI might be a good idea...and it might not. I wouldn't argue either
way. Among this community, good ideas *tend* to rise to the top.
Since this kernel is a work-in-progress, there's no implied guarantee
that interfaces and structures won't change. Now, if you think things
are being changed whimsically... %^)

Cheers.

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/