First, I need to acknowledge that I am quite impressed with RosettaNet's dedication to its current core competency: business process specifications to get trading partners to conduct e-business. When RosettaNet was launched, several key components like message packaging, routing, transport, business vocabularies, business process description languages, etc., did not exist. So RosettaNet built these components with the understanding that as soon as a suitable standard becomes available, RosettaNet would utilize that standard instead. For example, currently RosettaNet is looking at adopting the Business Process Specification Schema (BPSS), the ebXML Messaging Service (ebMS), and the ebXML Registry / Repository.
Conceptually, RosettaNet should be able to focus on the business aspects of the business process and express it in any of the business process languages (like BPSS, BPEL4WS etc). In reality however, BPEL4WS is simply not robust enough in its current form, to be adopted by a business-centric process specification organization. The origins of BPEL4WS are from the server workflow and application integration areas. It is important to note that the needs of trading partners wishing to engage in legally binding business-centric processes are completely different from the processes required to integrate applications, or internal workflows. The current BPEL4WS does not support the business semantic needs of B2B trading, and this is why RosettaNet cannot express its PIPs (ebusiness process specifications) in BPEL4WS (Disclaimer: these are only my opinions and not any views expressed by RosettaNet).
Currently, RosettaNet is the only concrete and effective XML-based B2B framework (100s of implementations, billions of dollars worth of transactions), and I think it should have the clout to engage BPEL to work together and make BPEL more robust for B2B. Only when this happens will RosettaNet be technically able to support the expression of its PIPs in multiple business process languages.