> I have just realised that nowhere in the report does it say specifically
> what the kernel language is. It says how to translate to it but not what it
The short answer is that the report defines an informal semantics with
a notional "easily understood", but imprecise kernel. If you want to
understand the semantics at a formal level, read the static and dynamic
semantics. Together, these define the complete language rather than
relying on implicit translation to a kernel (there is some explicit
translation in the dynamic semantics). Having worked with SML
semantics, I prefer this more direct approach to a translation: the
meaning (and type) of a source expression is rather clearer this way.
Both semantics are available from ftp.dcs.glasgow.ac.uk (and probably
animal.cs.chalmers.se and nebula.cs.yale.edu) in:
These are binary, DVI files, of course.