We use UNIQUEIDENTIFIERs for our PK values. We can generate them on the client when a new record is added. Therefore, we aready know what it is when we want to insert child records on the client.

>As I am working on my first SQL Server database, I came across the following question. The book "Client/Server Applications with VFP and SQL Server", in chapter 1 "Introduction to ClientServer" has a sentence:
>
>"If you have the choice, you should avoid identity columns when you design a database."
>
>I have a choice but don't know what to do :-)
>
>As I am playing with the database I created where tables have identity columns, I like it, since I don't have to do anything to maintain the PK value.
>
>In my VFP-database applications I usually maintain PK values for each table using a table PK_FILE.DBF where PK values get incremented by the code in the application. I suppose I can do the same in SQL Server but is it really an "accepted" way of maintaining PKs?
>
>What would you recommend to use for identity column/pk field?
>
>Thank you for any suggestions.