Okay, I've got good news and bad news.

The good news is that this will mostly work, and I can add several things to the list of stuff I no longer have to worry about.

The bad news is I have to add new code to any module that needs to load MIBs. The new code is as follows:

@DIRS = `ls -1d /usr/local/share/snmp/mibs/TEMP/*`;
foreach $dir ( @DIRS ) {
        chomp $dir;
        &SNMP::addMibDirs("$dir");
}
[download]

Anyway, some more detail on the advantages:

1. I don't have to modify environment variables. This is good because if $ENV{MIBDIRS} is previously unset, you have to find out where the site-wide MIBs are stored. While this is usually predictable, I'd rather not worry about it at all.
2. I don't have to modify environment variables. By adding the package's own base MIB directory to the config file, the change can be made effective site-wide (using the config file in /etc) without worrying about users' various shells, changing their environments, etc.
3. I can actually eliminate a config item I used to have in the code that kept track of the location of the site-wide MIBs that come with Net-SNMP. Again, less to worry about.
4. In the course of investigating this, I've discovered the proper use of the &SNMP::loadModules() function AND I've learned enough about duplicate MIB detection that the package's own code will be a lot leaner and the post-install script can do a lot more to detect duplicates (thus avoiding several really annoying MIB problems I've had before).

I may even write a separate script just to detect duplicates, since the user has a nasty tendency to do that. :-)

Anyway, I want to thank all those who offered suggestions. Hopefully this will be the last time I have to worry about this. :-D

--Rhys