>You could have a login table - one row for each user who is logged in. What >you do is, when a user logs in, they take the first available record, place >their information in there and RLOCK() that record. To check who is logged in, >you just see if you can RLOCK() the record yourself - if you can then the user >is no longer logged in.

This works with NT as your server (which is what Tim is using) but it may not work for all NOS. It used to be Novell held locks for a TimeOut period or until the same user logged back in after a mishap. At one point in its lifecycle, Banyan Vines held all locks until the server was rebooted (not a very robust operating system).

Just throwing some thoughts out. I was spring cleaning.