[2692] in testers
Re: rsaix 7.7E: cleanup bug - maybe ...
daemon@ATHENA.MIT.EDU (cfields@MIT.EDU)
Tue Aug 9 22:33:30 1994
From: cfields@MIT.EDU
Date: Tue, 9 Aug 94 22:33:07 -0400
To: jawetzel@MIT.EDU
Cc: marc@MIT.EDU, yandros@MIT.EDU, testers@MIT.EDU
(I'm continuing this discussion on testers since it started here.
However, future bug reports should be going to bugs.)
> attach x11r5
> attach motif1.2
> attach x11
> attach motif
For this and other reasons, I am wondering whether it's not the case
that the fact that multiple lockers may attach on a single mountpoint
is a bug.
X11 still attaches x11r4 on /mit/x11. Should it now attach x11r5?
Probably. But when we change it (and motif) how many people will be
surprised? But that's a different bug. I guess it's a question of who
loses, and how, when things are changed. Developers can be surprised
by not getting what they expect, and spending time debugging code that
suddenly doesn't compile because X switched under them. Users can lose
by unexpected functionality changes. But if there is no such thing as
"x11" pointing to the current location, we have to do more work to
encourage users to evolve.
Common mountpoints are useful, but they have this kind of lossage. It
would probably be cleaner to do away with them, but then developers
need to have fancier makefiles or modify them significantly when they
want to use a different version of X.
===
> I don't think that the average user should have to cleanup after the
> previous user just because he was unlucky enough to sit down at a
> machine that wasn't idle long enough for the attachtab to be cleaned
> up.
In a busy cluster, if you want this kind of a cleanup, you have to pay
the price one way or another. Either you have to wait for the cleanup
before you log in, or you have to do it yourself after you log in. I
suspect most users don't need this cleanup. And in your case, it's
more efficient to target the specific filesystems in question rather
than detach the world.
> I think every user when he/she logs on should get a clean machine.
This is a nice ideal (though perhaps it isn't quite ideal), and we'd
certainly love to be able to provide it. Unfortunately, the only way
to be sure of a truly clean machine is to reinstall it after every
logout.
So, you can't have the ideal. Since you can't, you have to define a
compromise and draw the line somewhere in your
time/cleanliness,utility tradeoff. The current implementation is
where I chose to draw that line.
Craig
