Relevant email from Kent Blackburn as to the requriements

I see this a a doomed exercise! The $DATA areas are too small to be
utilized for these workflows. We are just re-inventing a model that does
not work on the OSG for LIGO. The Caltech cluster does support accessing
the Storage Element(SE)/SRM from the worker nodes (directly). Many OSG
sites do although there are different usage cases (dcache API, etc.) but
those are not expected to cause the workflow any issues at this time.

The SRM API has been tested by hand and can continue to be tested by
hand very easily (srmcopy) I don't see how this gets more complex when
used in a workflow. And as I noted in my first paragraph, srmcopy should
not be needed.

There will be some OSG sites where the data has to be staged from the SE
to the worker nodes, but we can worry about them after we get this
working. I am mostly interested in getting data prestaged to OSG SEs and
making those OSG sites look like the LIGO Data Grid (no data movement
necessary). This will make the LIGO scientist much more interested in
using Pegasus and using the OSG!