It can be done by making a shared queue of filenames.
Filenames are put in the queue during directory parse, then they are removed from the queue and put in each respective file-message’s metadata in the inbound proc. If you really want to, you can even convert your file data into a message format that holds the filename as well and then route that across sites. (Something like xml, mime, tar, simple prefixed header.)
I’ve seen this done by using a plain old text file for the queue; for integrity, access to the queue file was synchronized using flock.
