Howard Smith and Peter Fingar caused a lot of commotion in the BPM/Workflow community when they asserted that "Workflow is just a pi-process" in this paper
. You can follow the threads of that never ending debate on the W3C WS Choreography mailing list
and on the Workflow Research Forum
While the title is an attention grabber, I find that assertion to be useless, even if it is true. Howard and Peter make this assertion on the basis that BPML, which apparently takes inspiration from Pi-C, can successfully model the majority or all of these workflow patterns
, and it can model the workflow engine itself (nothing to get excited about, a workflow engine can model itself).
So, what does that buy me? Pi-C, being Turing-complete, is capable of describing almost anything, but not always the best choice. Unless one wants to go to the extent of considering the number "1" a process, almost always one will add additional modeling layers on top of a theoretical framework such as Pi-C, rendering the theory less useful in offering formal semantics. In theory, the entire ERP system is a Pi-process, if anyone will implement it that way is questionable.
Anyway, there are a lot of insightful posts in the threads above, for those interested.
posted by JC Reddy
: 12/21/2003 05:08:00 PM