Re: executable semantics, was Formatting of Language LRMs

Ivan Godard <ivan@ootbcomp.com> wrote:
> How would it differ from a document containing,
> > say, Java?
>
> If the document were a compiler written in Java (i.e. a recognizer and
> checker) then formally there is no difference. Practically, an immense
> difference. VWG is an ASL for specifying languages. All of Algol68 in
> VWG was about four pages of appendix; a compiler for Algol68 written in
> Java would be much more than that, even after you strip out optimization
> and code generation.
>
> Also, the specification in VWG lends itself to use of metatools: make a
> change to the grammar and verify that the revision is complete,
> consistent and not redundant. Not something easy to do when the language
> is expressed by a Java recognizer.


FWIW if I remember correctly a varient of that belief led the Modula II
standard to attempt to descibe the langauage using the Vienna Definition
Method. They hoped to verify the formal description automatically, but
they could not afford the commercial tools and the implementation of one
of the participants was not ready in time when they decided they had to
get it out the door.

Return to the comp.compilers page.
Search the comp.compilers archives again.