What about the marriage licence of a couple? Does it belong to to the husband, or the wife? Where should the marriage date be stored?

The Marriage table.

In my current project, the database design has undergone at least 7 revisions, several of them quite extensive. I think we have finally found the proper schema (at least I hope we have). In contrast, the object model I drew up at the time of the original database design has remained unchanged throughout the entire revision history of the database. Oh wait, I forget. We did change the name of one of the classes (without changing any of its behavior).

I cannot verify such anecdotes without seeing the details. I get emails from readers with anti-OO anecdotes also. A my-anecdote-can-beat-up-your-anecdote debate rarely goes anywhere.

I would like to see you apply quality analytical and articulation skills to describe what specificly is faulty about relational theory that resulted in so many alleged revisions. Database normalization is pretty much based on once-and-only-once (duplication factoring) at its root. Did OAOO fail? That would be very revealing if the case. Sometimes OO shops create too many tables. I suspect this is the culperate.

Further, if we don't document our lessons, then people may keep making the same mistake throughout history. Further, just because technnique X blew up on you when you tried it in 1983 does not mean that is the best way to go about using X. Expose it to more eyes and they may spot a variation that you did not think of at the time to clean it up. Isn't this how open-source is supposed to improve things also?