Richard--
Barton wants a total ordering over arbitrary expressions in order to
have a semi-canonical form for sets. This is a reasonable thing to
do -- in fact, the general simplifier already does it. The way it does
it, of course, is to use general representation (not Rat form) for
everything.
As I've discussed with him, the best you can do anyway is a
*syntactical* total ordering. There is no hope of a general total
ordering, since in general you can't find a canonical form for arbitrary
expressions.
Anyway, as usual, "1 is not of type LIST" is not a reasonable error
message ("1" doesn't even appear in the input), even if an error should
be given.
-s