One of the larger pains here is all the dependencies that DTK has. It is built on top of Trilinos and MOAB... and it uses a ton of stuff from Trilinos:
In particular... the Tpetra part is a pain. We currently only wrap Epetra with libMesh.... so a big part of getting this whole thing going is to grow NumericVector interfaces for Tpetra.... which is always fun.
Also, I am hoping to still be able to solve with PETSc.... and then copy in and out of NumericVectors (that are wrapping Tpetra vectors). So I'm going to be configuring libMesh with both Trilinos and PETSc support simultaneously... that should lead to a few fun issue ;-)
But... if DTK does what it claims to do... and does it well, it will be worth the pain....