>Still setting up my linux system and now I'm branching out and getting
>packages like PCMCIA, Netscape, isapnptools, and Samba. Is there a
>directory that should be used when unzipping a file? I've been putting them
>in /usr/src and then unzipping/untarring them from there with some very
>interesting results.
This is probably not the best way to go about it. As you've seen,
there's no de facto standard for how packages are distributed.
The standard formats that exist (GNU autoconf for source, Debian
and Red Hat for binaries) are excellent, and will eventually
become ubiquitous, I think.
But if you have some non-brand-name :) binary distribution,
it's best to unpack it in a separate, clean directory, then
manually put all the files in their proper places.
Alternatively, view the archive (tar -tzvf or unzip -l)
to see what it will do when extracted, and position
yourself apppropriately before extracting.
Quote:>Netscape made
>one or two directories under /usr/src, but it also dumped a bunch of files
>in /usr/src.
On my carefully-organized system, all of Netscape is contained
in its own '.../crappy-setup' directory, with appropriate
links from where the files *should* have gone. I fear Netscape
is still thinking it's on a Windows system, where nobody
really cares if an archive barfs files all over the filesystem.
(yeah, so that was a bit uncalled-for. But I feel *better* for
having said it:)
Quote:>I understand the concept of 'it's MY OS and
>I'll put files anywhere I want them to be', but for someone just starting
>out a general rule would help.
The File System Standard (FSSTAND) is pretty good, and used
intelligently by Debian, and probably Red Hat.
In my irregular opinion:
/sbin files needed for emergency/boot, only used by root
/bin other such files, useful to everyone
/etc system-wide config files, system scripts
/lib Really Important (tm) libraries.
/usr/* Same classifications not less critical stuff.
/usr/local* Hack packages.
I think the current setup is something of a mess. I want my system
to keep packages seaparated by name, version, and architecture.
This lets me 1) see what package a file belongs to 2) keep
multiple versions of a package on the system, 3) cleanly
destroy unwanted packages, and cleanly move packages from
system to system as tarball, and 4) install packages that
have conflicting files without one overwriting the other.
I have (almost) a complete, bootable, glibc-2 system that
is set up this way, where /usr is a pure symlink tree.
Eric