tytso at mit.edu
Thu Jan 3 16:37:14 PST 2002
On Fri, Jan 04, 2002 at 12:34:18AM +0000, Alan Cox wrote:
> > Glibc already contains so much version checking code that it would be no
> > problem to use /proc/cpuinfo if the kernel version is smaller than
> > foo.bar.baz and a sane interface if later.
> Wake up, hello, anyone home ?
> The _current_ glibc doesn't know about this, so you will break existing
> systems. Unless you can grasp that elementary detail you aren't going to
> have a userbase.
Absolutely --- Alan is 100% correct here. The earliest /proc/cpuinfo
could disappear is 2.8 (assuming it was deprecated in 2.6). And I'm
not convinced it's non-pid /proc entries are so horrible that they
have to go away at all costs --- there are far more important things
for us to worry about in the kernel at the moment than aesthetic
issues like this.
Sometimes, practicality needs to take a back seat to "the Right
Thing". Remember, sometimes "Worse is Better"
(http://www.dreamsongs.com/WIB.html); in fact, many people would argue
that Linux is a prime example of how sometimes "Worse is Better"
approach is far better than trying to do the Right Thing. (That way
lies NetBSD --- the organization that took three years to rearchitect
their entire device driver interface before they had full PCMCIA
More often than not, practicality and support of the existing user
base is far more important that being Architecturally Pure.
More information about the lsb-discuss