Michael zur Muehlen just posted an paper on The Business Process Analytics Schema:
http://www.bpm-research.com/2009/02/22/the-business-process-analytics-format-bpaf/
The state-oriented appoach makes perfect sense to me. It is a straight forward shema design you can easily incorperate into your own process exception handling framework. You can also extend it by adding more fine-grained sub-states for your process minning needs.
There is a minor issue I think can be improved to make it more elegant mapping between this schema and its object oriented representation. Here is the exact comment I left to Michael:
I think it’s better to move bpaf:DataElement node under an additional node called bpaf:DataElements - an explicit collection-type node. This way, we will be able to map this schema to a object model elegantly. You can easily model a class hierarchy representing that an Event class has a collection-type memeber called DataElements which consists of DataElement-type items. You can’t directly model an Event class that has an ambiguous number of DataElement described by maxOccurs=”unbounded” attribute. I believe this will ease the adoption of this standard in practice.