PMFJI, here's my $.02 worth.

>>Well, it could still be a very real problem, couldn't it. . . what if someone goes to lunch in edit mode, deliberately expecting to "hold" the record until they return?... or to a rush meeting?... or a mad rush to the john?...or has to look something up in some not-handy reference... or ...?

I use pessimistic row buffering, w/ EDIT and SAVE buttons and a timer. If the user starts to edit something and there is no activity for x minutes, the change is cancelled, the user notified and the record is unlocked. Seems fair to me.

I have always thought that it would be rude to allow a user to start editing a record, THEN after he's entered/updated data, tell him someone else made a change before he could save the data.

I work mostly w/systems w/only a few users.

I've recently read a lot about optimistic buffering being preferable, but have not been convinced...maybe it's just my 'old fashioned' way of looking at things.

I then use transactions when I have changes to make to more than one table.

Great info/discussions here!