>>Thanks for the response. The reason I asked Bonnie is because she suggested earlier that change tracking should not be handled in a partial class but at the parent tier. Hence my question. ;-) To be honest I can see justifications either way.
>
>From the perspective of the referenced article, it seemed to refer to business type "behaviour" being absent from the typed dataset (as opposed to change tracking, per-se). My observation was purely on that basis; in otherwords, behaviour can be "encapsulated" with a dataset if that's how you want to handle things.

I can see some justification for the practice of calling services from nearly empty BO's when using single inheritance languages: aggregation and delegation into aggregation are (besides interfaces) another way to implement behaviours otherwise "gained" via multiple inheritance. Coupled with a factory approach you get a lot of functionality with little code almost as polymorph as in multiple inheritance. But that might lead to arguements on how qualified the analysis of the biz task really was<g>.

my 0.02 EUR

thomas