Linus Torvalds <torvalds@transmeta.com> writes:
> That code should just DIE.
This was an example. The STREAMS stuff is simply something which came
immediately to my mind since I know there are parameter which must be
exported by sysconf.
> Code does NOT get added to the kernel just because somebody makes a big
> deal of nothing.
This leads to problems. People will be surprised to seem some
programs run slower in future because some of the few remaining fixed
limits in glibc are lifted (because there are people who recompiled
their kernels) and the only way to accomodate this is by trial and
error. In some cases hundreds of syscalls could be made because the
information is not available. And note that even if somebody would
volunteer to work on implementing this for the kernel right away there
are still old kernel versions to take care of.
By thinking a bit ahead and investing a few more bytes (the sysctl
stuff really requires not many resources) you can avoid this.
Otherwise we'll have ever more configuration parameters in the libc
since every single kernel version must be handled specially. This
descreases performance even more since the duplicated code&data is at
user level where sharing (of the data) is not possible to that extend
and where the kernel differences have normally to be handled
dynamically instead of at compile-time.
I'll stop trying to convince you since I don't have much hope. When
glibc 2.2 comes out I'll provide some information on how much code is
necessary to handle all the different kernel versions. Almost all of
these changes could have been avoided if this purely minimalistic
approach to kernel interface design would be replaced by something
more flexible.
-- ---------------. ,-. 1325 Chesapeake Terrace Ulrich Drepper \ ,-------------------' \ Sunnyvale, CA 94089 USA Red Hat `--' drepper at redhat.com `------------------------- 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/
This archive was generated by hypermail 2b29 : Mon Jul 31 2000 - 21:00:24 EST