unique id for data types

[Barney Hilken]

Ok, I'm stuck.

The Oleg stuff is great, and it looks like he did everything I've  
done long ago, without needing type functions or associated types.  
But I can't use the TTypeable stuff directly because you can't  
convert the "relational" style of Oleg's work (using functional  
dependencies) to the "functional" style I'm using (using type  
families). Rewriting all the basics in my style is no problem  
(especially as I only need a limited part of it) EXCEPT Derive  
depends on (at least the types of) Template Haskell, and TH can't  
define type family or associated type instances.

So I'm back where I started.

Barney.