I haven't followed the discussion closely, but to throw out another scenario where the current procedure semantics seem to fall short (assuming I'm not overlooking something):
I recently turned a series of functional test cases into DocBook. Each case was 1 or more actions, followed by 1 or more expected responses, possibly followed by more actions and responses. -- John R. Daily <email><mailbox>john</mailbox><domain>geekhavoc.com</domain></email>
