>Hi Keith
>
>I would guess that it depends a lot on what can be assumed about the data source(s). One would expect that within the source, the uniqueness of primary and candidate keys should always be met.

...such is the nature of primary & candidate keys...

> Adding a field for the source and treating it as part of the merged primary key should then be easy, and errors should be rare, so handling the error should be more efficient than a seek for every record to be merged.

This is a recurring merge where data from a single source
will occur on a fairly regular basis. Some of the sources
send us updates only, others send full files each time.

At this point I can make no assumptions about the hit rate.
My question in essance is: is processing an irregularly
occurring error more "expensive" than doing a SEEK before
each add or change?

   ...kt