Re: Semantics of Internal Definitions

[John D. Ramsdell <ramsdell@linus.mitre.org>]

> From: jar@cs.cornell.edu (Jonathan Rees)

> But... but... there is a difference between unspecified and
> unassigned.  The translation you give makes this valid:
> 
>     (let ()
>       ((lambda ignore 3) x)
>       (define x 7)
>       x)

I think the semantics of internal definitions should be specified only
in the case in which all variables are defined before being referenced
or assigned.  In the example you give, since it does not meet the
condition, an implementation may compute the number 7 or 99!  An
implementation is free to do what it pleases.  In summary, I think the
semantics of internal definitions should be defined as in the first
appendix with the same preconditions given for toplevel definitions.

John