Interesting take on this topic from Keith Harrison-Broninski over at BP Trends (PDF download). This is the fourth part of an ongoing series on the future of BPM. You can find the earlier entries in the series here.
Keith asserts that BPMN will make BPEL unnecessary. I've always thought about BPMN a sort of XMI for sharing a processes model between design tools instead of an interoperable model for instantiation and execution. Maybe I need to re-think my opinion on this. I was surprised Keith didn't mention of the 1:1 mapping between BPMN and XPDL 2.0. This avoids the lossy transformation of moving a BPMN model into a BPEL serialization, since BPEL is effectively a subset of what can be represented by BPMN (and most integration tools).
The most interesting part of Keith's article is his discussion of orchestration and choreography. According to Keith, orchestrations describe what the overall process appears to do without specifying how any of this is to be implemented. This sounds vaguely to me like abstract BPEL, a concept that, unfortunately, never seemed to catch on with many people (its also where the value of BPEL can be realized).
The biggest challenge to a choreography is implementation. Since a choreography is a form of peer-to-peer interaction there is no "conductor" - this means that the choreography is an agreed-upon model for interactions that consists of a series of orchestrations (see the simple slide I posted yesterday as an example).
Will BPMN make BPEL irrelevant? It remains to be seen, Most of the vendors and people I talk to are interested in converting their BPMN process models into BPEL for instantiation within their favorite integration platform. Keith's next installment promises to further examine business process standards, orchestration and choreography. Sounds like an interesting read...
In the meantime the OASIS BPEL Technical Committee will be in Los Alamos next week for what might be one of our final face to face meetings prior to standardization. The feedback from the Public Review has been fairly calm and we've got a few outstanding issues to get resolved prior to pushing ahead with the standardization process. Given the Christmas and holiday season I expect that standardization won't happen until January 2007.
Dave Linthicum is blogging about orchestration and gets it mostly right.    The blog entry