Excerpts from transarc.external.info-afs: 16-Feb-93 Re: ls /afs: It
takes a lon.. John [EMAIL PROTECTED] (608)
> In any case, there is already
> local abbreviation of pathnames. Look in root.afs in the athena.mit.edu cell
> or the cs.cmu cell. If I say "/afs/athena/user" to someone at a remote site,
> chances are that user has to change the name to obey the local conventions
> for root.afs. Users at CMU say "/afs/cs", but we don't have such a directory
> here at MIT. The battle for unique global pathnames is already lost.
Nope (and I assume you meant cs.cmu.edu, not cs.cmu). You should be
able to canonicalize any name into a global one, and you should even be
able to do it programmatically, irrespective of the local abbreviations.
If there's an entry in any root.afs that isn't a real cell name, I'd
hope that it is a symlink to the real cell name so that it's obvious to
all agents, including programs, that the shorter name is just a
nickname. Either people or programs can then do the canonicalization,
and it's simple to ask that all transmitted names be canonicalized.
(It's just thoughtless not to do that.)
The idea is that there should be a canonical global name for any AFS
file, and it should work from anywhere that has access to the cell that
stores the file. There are lots of aliases, not just the symlinks
stored in root.afs; lots of workstations have things like /usr/bin or
/usr/local/bin as symlinks into AFS somewhere, often pathnames with
/@sys/ in them somewhere. Sure, I can think that it makes sense to tell
somebody about the neat new /usr/local/bin/dog_breath program, but I
should have sense to try (cd /usr/local/bin; pwd) first if I really want
to point somebody else at an AFS file.
Craig
