Another View on BPEL4People

Richard Brown, an IBMer from across the pond, blogged recently about BPEL4People and took issue with my contention that it was too grandiose and after-the-fact to achieve wide adoption as a standard. He tracked back to my original post in February, saying

I'm increasingly of the opinion that standardisation often occurs too soon and that major revisions are a reflection that the initial specs fail to anticipate potential problems or extended use cases. Unfortunately, if vendors choose to delay standardisation, they're accused of being proprietary or risk finding themselves with no influence amongst those who decide to standardise earlier. When the incentives are so strongly stacked in favour of early standardisation, it's not surprising that those who gain the most experience with a spec discover its deficiencies and seek to remedy them.
Anyway, since my original post was from February (before I even had my own blog), I'm hoping for a restart on the discussion of BPEL4People. What's the right way to include human tasks in a BPEL process, and why?