I know, I'm late. I've been meaning to read this all week, but I kept putting it off.
OK, here it goes.
I think Andy has done a great job pointing out some things. I totally agree that something is better than nothing and if you have no clue, then make an identity field and add a clustered index. No matter what. I thought I had a reference that shows where a clustered index is better than a heap in terms of performance, but I can't find it. If I do I'll post it.
Not having a primary key is a sin. Not having an intelligent PK is not as bad, but doesn't excuse the developer. Andy's point to this article (I think ), is that this is a Worst Practice, not a recommended practice. He is showing the lowest level of something that should be done.
Steve Jones