On Tue, Dec 19, 2006 at 07:07:52PM +0700, Ali Milis wrote: > Hi, > > Suppose there exists large-company.com and > branch1.large-company.com. Both are producing "local packages" > which may conflict with each others. > > Question: > Is it OK to use /opt/com/large-company/ and > /opt/com/large-company/branch1/ for local packages? > Or should /usr/local/com/large-company/ and > /usr/local/com/large-company/branch1/ be used? > > My understanding of FHS is that /usr/local/ is trully > for local host. > The FHS is quite flexible in this respect. Personally, I like /usr/local for stuff that I compile myself. For instance, I use a version of qemu that I compile myself (which I install into /usr/local/stow/qemu-<version> and then use stow to manage, but that is beside the point).
Stuff that goes in /opt includes monolithic apps which I get in binary
form. For me, that includes Eclipse, the upstream version of {hot-or-
cold-substance}{small-mammal} if I need it, LTSP and other such things.
Those things end up installed in directory trees that look like
/opt/eclipse-3.2, /opt/firefox-2.0, /opt/ltsp and such. Then I
generally create a script in /opt/bin which launches the app. For
Eclipse, I have:
#!/bin/sh
cd /opt/eclipse-3.2
export LD_LIBRARY_PATH=/usr/lib/jni
export LANG=C
./eclipse $*
That suits my needs quite well. Now, as long as the amount of stuff I
have hanging around in /opt is not *huge* (and by huge I mean 100+
packages) I like to keep things in the first level of /opt named for
their packages, with the exception of /opt/bin and /opt/man (for
programs which include man pages I would like to have available.
Regards,
-Roberto
--
Roberto C. Sanchez
http://people.connexer.com/~roberto
http://www.connexer.com
signature.asc
Description: Digital signature
