Re: containers/chroot to allow ABI breakage is the wrong approach
On Fri, 24 Oct 2014, Thomas Goirand wrote:
> > OpenBSD’s libc.so major number is 50 or something like that right now,
> > because they – correctly – increment it on every incompatible change.
>
> The correct thing to do is to not do incompatible change.
No, in the interest of software hygiene it is often useful
and sometimes even required. I can think of cases where
glibc’s binary compatibility is *still* biting users’ arses.
lewellyn and I had an enlightening discussion on IRC yesterday,
which touches this issue. I won’t bore you with the why and the
details, but imagine a musl/Linux distribution with the Unix
concept of “universes”, which offers a GNU/glibc universe,
specifically for running proprietary/binary-only GNU/Linux binaries.
The irony!
> > This is not a problem because, you know, we have Open Source, so we
> > can always just recompile everything against the new libraries.
>
> Wouldn't it be better to "just" upgrade to the new lib? Recompiling is a
> major pain and a loss of time/resources which could be avoided.
Sometimes, yes. Often, though, the pain is manageable, and if you
do it often enough you have incentive to keep the resources and
especially manual work required small.
> I explained extensively why in my post. Re-read it, and let me know
> which part you didn't understand... :)
You wrote a whole lot of things I fully agree with, but you
postulate that library major bumps must not happen, without
explanation or reasoning, which I disagree with based on my
current understanding of things.
bye,
//mirabilos
--
Sometimes they [people] care too much: pretty printers [and syntax highligh-
ting, d.A.] mechanically produce pretty output that accentuates irrelevant
detail in the program, which is as sensible as putting all the prepositions
in English text in bold font. -- Rob Pike in "Notes on Programming in C"
Reply to: