version for Patrick's review

This version is for Patrick to review.
It differs from the rough version in that some obvious problems have been 
fixed, and there is a change to the abstract syntax permitting a graph to 
have more than one name, (and not only equivalent graphs to have one name 
each).

Also the introductory text is getting longer, and the references are 
filling out.

Chris and I are still working on the soft text (particularly sections 2 and 
7) -  we need to work in a bit more of Chris's earlier work presented in 
Cannes, the use cases etc.

I will post another version in about six hours time, with the latex source.

Jeremy