Depending on the complexity of the filtering I lean towards keeping it in the Xlate. If you have multiple and/or compound requirements it usually comes out cleaner and easier to develop using Tcl, at least for me. I usually resort to Tcl if the Xlate method will require more than 5-10 additional actions. This is an attempt to reduce Xlate bloat which I find harder to support than reading Tcl code.
I have developed several interfaces like this and neither approach is easy to support as you get into more complex filtering. I think the question you should answer is which are you more comfortable developing in Xlates or Tcl? Also, from a support standpoint how comfortable is the rest of your team with the approach you choose. Meaning, don’t pick Tcl if you know it, but nobody else does.
