>You beat me to the point that strong typing does prevent subtle bugs. If you are trying to do something that is inappropriate for one or more of the operands, you find out at compile time instead of run time.

Very true. Strong typing (and specifically typed datasets, which started this discussion) have benefits of discovery with Intellisense, potentially requires less conversions, the schema generation tool will generate methods to automatically traverse hierarchical data, they serve as a form of self-documentation for stored proc result sets (which is possibly HUGE in multi-developer projects), they make design-time data binding easier, and will potentially help many large reporting projects by allowing developers to design report content even before the stored proc/query has been built. Their ability to serve as a form of documentation for the report data source is quite helpful, especially in multi-developer projects.