Hi, > But how is the vendor supposed to know that GNU libc6 requires the > files to be oriented according to the FHS? That should be in the > glibc docs, period. Who said glibc requires the files be arranged a certain way according to FHS? The original statement I responded to a few posts ago was: GW>> In a modern GNU/Linux distribution, /usr/include/linux should GW>> *not* be a symlink to anything at all. If glibc really requires things a certain way, that's not a very flexible package. I don't think we proved that yet (though I admittedly follow the threads loosely). The distribution vendors know what to do because they have a person or team which subscribe to the Standards orgs and go over the guideline docs for changes which conflict with their current way of doing things. Looking at the membership list of the Free Standards Group (which promotes LSB and FHS), you'll see both RedHat and SuSE (among others) are Gold members. You typically don't invest to become a Gold member of some organization unless you're serious about it. Consider that their recent distributions have been LSB certified...so obviously they found out somewhere sometime how things are supposed to be laid out. Indeed, the /usr/include/linux and /usr/include/asm on my RedHat AS/ES/WS 3.0 machines are actual directories and not symlinks as used to be done in the Good Ol' Days. It would be nice for glibc to give a Release Note about the change in the symlinks, but to say it *must* be there relies on three things: 1) that the Standard came out before the glibc package was released. 2) that the chosen Standard is The One and Only Standard that doesn't ever compete with some alternate FooStandard 3) that glibc is a linux-only package and has some specific tie to Linux kernel headers My intent in mentioning LSB/FHS in my first post was to not really take any side of this particular argument, but to point out that they exist to attempt to bring some order to the chaos that's been Linux distribution layouts. Once that order is adopted by the big fish in the Linux world, any workarounds cdrtools or others have employed should be unnecessary. I don't agree or disagree what's the proper way to do things, but I do take exception to some commonly seen viewpoints on this mailing list that Linux will forever remain an unorderly pile of goo compared to say, oh, Solaris. -Robert
Attachment:
pgpPuNbVyj_gU.pgp
Description: PGP signature