Erik-
No, I *shouldn't* have to. While I don't do as much graceful error correction as I'd like, I do at least always try to account for really fatal things like a primary key failure. So, I think of it as a redundant safety system...one that's actually getting used < s >.

As for how I gen my keys. Well, I was afraid you were going to ask me that! I'll tell you if you promise I won't get any flames for it < bg >. Seriously, the primary keys for all my tables are character (icky, I know) based on a combination of date/time stamp (useful info for audit trails) and, because of a glitch with NT and SECONDS(), a random key. I had lot's of trouble with id collisions on NT before I put in the random key when I added lots of records quickly on a fast machine.

For the first day I had my current problem I assumed that I was just colliding/overstepping the ids, but everything I tried to narrow that down came up zilch. So, I spent a few more days hunting down what turned out to be phantoms.

< Shrug > Probably more than you wanted to know...but that's the scoop.

>You still shouldn't have to do that. How are you assigning the pk?
>