For the other two people in the world who are interested in such things, Peter Lin, Chimezie Ogbuji, Johan Lindberg and I have been debating the finer points of Rete engine implementation over on Peter's web site.   The main thing to emerge from this was a much clearer understanding of the alternative approaches that can be taken when implementing Rete networks (I told you there would only be a couple of people interested in this!).   Chimezie is the author of a reasoning system called FuXi (pronounced 'foo-see', please) which extends a Python-based engine called Pychinko.   Pychinko and FuXi are designed to apply rules to RDF triplets, and therefore can be applied to the semantic web.   With good reason, they use a rather different implementation for the alpha part of their Retes than the approach used in most engines. 

All a little obscure, I realise, for my normal readers, but someone may be interested.

The debate can be found at:

Difference between RETE/UL and Forgy RETE
http://woolfel.blogspot.com/2006/10/difference-between-reteul-and-forgy.html

For me, the debate began earlier at:

Wikipedia has been updated
http://woolfel.blogspot.com/2006/10/wikipedia-has-been-updated.html

Previously posted at http://blog.solidsoft.com/blogs/charles_young_mvp_biztalk/default.aspx