ekrell@hector.UUCP (Eduardo Krell) (11/09/89)
In article <4555@emory.mathcs.emory.edu> arnold@emory.UUCP (Arnold D. Robbins {EUCC}) writes: >Of course, the key words in the quote are "Lacking a more formal definition..." >It is to be strongly hoped that the C++ Reference Manual which has just >become available will be the final arbiter of C++ semantics, not "whatever >cfront does". I find it hard to believe that a document written in a natural language will provide a complete, unambiguous semantic definition of a programming language. The last time that was tried was Ada and even though they intended the Ada Language Reference Manual to become the ultimate semantic definition, they still needed a committee to resolve ambiguities and inconsistencies within the LRM and the interaction among the most obscure Ada features. Eduardo Krell AT&T Bell Laboratories, Murray Hill, NJ UUCP: {att,decvax,ucbvax}!ulysses!ekrell Internet: ekrell@ulysses.att.com
jacob@gore.com (Jacob Gore) (11/09/89)
/ comp.lang.c++ / ekrell@hector.UUCP (Eduardo Krell) / Nov 8, 1989 /
I find it hard to believe that a document written in a natural language
will provide a complete, unambiguous semantic definition of a programming
language. The last time that was tried was Ada and even though they
intended the Ada Language Reference Manual to become the ultimate
semantic definition, they still needed a committee to resolve ambiguities
and inconsistencies within the LRM and the interaction among the most obscure
Ada features.
----------
I don't think you can avoid having unforeseen questions about the
definition, whether it is "formal", "natural language", or "operational" --
especially about the "interaction among the most obscure features".
Sure, a natural language description will have ambiguities that are
inherent to natural languages. But this only covers the cases where the
authors of the document intended to provide the information in the first
place, and it just came out ambiguously. Omissions are a different matter.
The Algol-68 Revised Report[1] can hardly be described as "written in a
natural language" (certainly nothing used on this planet:-) -- but that is
no guarantee against things being left out of it. (If that was the case,
there wouldn't be a need for a Revised Report, would there?:-)
And the same problem occurs with using a reference implementation: to be
executable, it has to be written for a specific set of platforms, and it
just doesn't address questions that will show up on other platforms.
------
[1] Revised Report on the Algorithmic Language Algol 68.
A. van Wijngaarden, et al, ed.
Springer-Verlag, Berlin, 1976
--
Jacob Gore Jacob@Gore.Com boulder!gore!jacob