Jim,

I understand that this approach does serve you well Jim, but on the other hand I think it's wrong to FORCE everyone to use generated primary keys. After our last conversation regarding this subject I did discover some problems that wouldn't have arised when using your approach. But saying that PKs should have no meaning outside the database is too farfetched to me. Virtually none of the learning books which handles the relational model do emphasize the use of generated keys. IMO it violates it to some extend.

Walter,

>
>I stand by my statement. In relational database theory the single purpose of a primary key is to uniquely identify a record within a table. If it weren't for that purpose then the concept of primary key would not exist at all. Primary key is defined as, "An attribute or set of attributes that uniquely identifies a record within a relational table."
>
>With that definition and the fact that only relational database systems require a primary key, I still say that the unique responsibility of the primary key over any other attributes is to uniquely identify a record so the computer can find that record unambigously. I also say that the best primary key is the one that does that job best, fastest, and with the least amount of maintenance required. The approach that meets these criteria is an integer sequential computer assigned primary key and the relagation of all other alternates to be alternate keys only.
>
>You and I and anyone else can argue this point to the death, however I have used this approach and it has served me well. I do not have miriads of RI code for updates because the PK never changes value. Users are happy because they can feely change the value of any field they see and work with at will. The occasional situation where someone will open one of my table without using my application has never caused a problem because, I believe, that the PK field is totally meaningless to that person and they therefore don't modify it. My approach with VFP data has translated very easily to using identity fields in SQL Server. My development is a bit easier because every single table has the same type of PK and the values are generated the same way