I see the former as reinventing the wheel rather than the latter, but I can see reasonable people on either side.

Incidentally, a lot of stuff in SSIS can be used as a crutch as opposed to bringing in the most final/discrete dataset from the initial SQL pull. Same thing can be done with Crystal Reports, but my preference is to get the job done as best as possible in the initial data pull.

There are numerous reasons for this, but one relevant one that comes to mind at this moment is ease of being found by another developer.

I hate going into an application and looking in 16 different places to find out why something is happening, when all 16 of them could have been done in a single place..