group membership and protection

[utzoo!decvax!ucbvax!unix-wizards]

>From decvax!duke!bcw at Berkeley Tue Oct 27 19:54:35 1981
It seems to me that it is reasonable for one user to
belong to multiple groups;  we have a locally-written
data management system which uses a similar scheme and
have found it to be reasonably close to what we need
for solving problems with multi-user access.

This scheme is not as "overhead-bound" as some of the
contributors seem to think.  It is quite easy to keep
a small table in memory for each user keeping the n
most recently used groups.  Since access tends to be
fairly localized, this gives reasonable performance if
n is even as small as 3 (at least in our environment;
although it is possible that there exist places which
might have a problem).  If this number isn't sufficient
it is easy to increase it to, say, 10 or any higher
number to obtain reasonable access time.

Also, since the list is kept sorted in memory, you
usually get a hit within the first or second group and
aren't doing an unbounded search through the list.

If anyone sees any problems with this scheme I would be
interested to hear about them.

				Bruce Wright (Duke)