On Sun, 29 Jun 1997, David Boyes wrote:
> On Sun, 29 Jun 1997, Norm Aleks wrote:
> > You're making things complicated by accounting only by file ownership.
> > Why not just make a special case for LISTSERV?
>
> This would work, but is the beginning of a very slippery slide if you're
> running a very large Unix environment with multiple thousands of users. If
> you're really serious about charging for Unix usage, about the only real
> mechanism to defeat clever twits is to charge by file ownership. Otherwise
> you get random twits storing huge files in unused accounts or other such
> nonsense. Sure you can code this in, but what happens with the next
> application that runs as root or another user and creates files that you
> want to bill to a specific user? You end up maintaining a whole list of
> special cases, which makes the accounting code difficult to maintain.

You've said it better than I could.  We used to have a lot of special
charging cases on VM and MVS, and it was a nitemare to support, so we've
worked for many years towards simplifying resource accounting, which has
meant sticking to the accounting interfaces in each opsys.  In this case,
we use standard Unix disk accounting, which charges files to the owning
user (UID in the inode).  We don't have to create special accounting
schemes for other Unix data-storage servers such as personal email or ftp
or web servers, so why should we have to charge specially for LISTSERV?
Also, the Unix sysadmins we can find nowadays don't have a clue about how
to interface to MVS where our central accounting system resides, so adding
another charging scheme for LISTSERV would take much more work than just
writing some simple Unix programs.

David R Nessl  -- Coordinator, Computer Systems (sysprog/sysadmin)
http://www.nerdc.ufl.edu/~david