[Sml-implementers] real constant too large

[Stephen W. <sw...@sw...>]

Does anyone have any thoughts on what an SML compiler should do when
it encounters a real constant that is too large to be represented?
Currently, some compilers silently produce an inf (Hamlet and MLton),
some report an error (Moscow ML), some encounter an exception in the
compiler (Poly/ML and SML/NJ), and one reports an error during
assembly (ML Kit).

My thought is that the right thing is to do as Moscow ML does, and
report an error similar to an integer too large error.