Tue Mar 25 14:43:30 EDT 2008

calling transformers directly

isn't really good style. why? they all run in the same lexical context
as the transformer they are called by.

need to have a better look at local-expand and friends to see if
there's no better way to handle this. in other words: check where is
it (not) necessary to have the same lexical context.

i guess it's ok for the RPN code representation: a piece of RPN code
produces a single lambda expression, which is a single lexical
environment anyway..

but i don't see if it's always harmless.
maybe i should just try to break it?

because of the way that early expressions are on the deepest nesting
levels, introducing new names into the expansion only influences code
BEFORE a certain point, so i see no reason to do so.