Use of identity columns (or any other contrived key) certainly has it's foes - and I'll agree that each side has it's proponents. What I tried to do..and evidently failed...was to target not users like yourself who have some experience in this arena, but to point out to users who DON'T know, understand, etc...that failing to use a primarykey is horrible. Should I explain the differing views on contrived keys? While maybe not doing so left a gap in the article, my thought was (and is) that for developers/DBA's making the worst mistakes don't try to change it all in a day, try to get them moving along the path to doing things better ways. If you're not seeing worst practices, you're working in a better environment than I do!

Clustered keys now, I'll have to disagree. Platform independence is the holy grail I think, and seldom reached. Clustered keys are a MAJOR part of SQL, to not use them....I just don't see that. In your example, I'd look for a better column to use first - often a date column, but in any case one where I'd want to return the results sorted. Failing to find that, I'd probably use the primary key as the clustered key. The only time I've had issues with contention on clustered indexes was when they were frequently being updated. I'd encourage you to revisit this issue, you may find performance gains.

Thanks for your feedback, I hope you'll continue to participate and offer your opinions - as you can see from this discussion, you're not the only one who disagrees with me!

Andy