[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Scheme-reports] multiple values module

Andre van Tonder scripsit:

> I prefer William Clinger's philosophy regarding standards, which can
> be found somewhere in the R6RS discussions.  From memory, a standard
> posits a set of postulates that can be used to infer a set of portable
> programs.  Not all programs allowed by an implementation need to be
> portable, but if the set of postulates is simple enough, it should be
> relatively easy to program in the portable sublanguage.

That sets a floor, but not a ceiling.  How much should be standardized
and how much should not remains open.  In ANSI C discussions, the
formula that a standard is a contract between users and implementers
was employed.  This too is reasonable, but how many clauses should the
contract have?  C prescribes that a short is no shorter than a char and
no longer than an int.  Java prescribes that it is 16 bits.  Both styles
have merit in their own contexts.

> Specifying that an unspecified value be returned by a command is a
> prime example of a a postulate that does not need to be added to the
> language specification.

But it *was* added and remains there.  To change it, a majority of a
self-selected set (you, too, could have participated, and you and you
and you) had to decide to change it.  They didn't.

All Norstrilians knew that humor was            John Cowan
"pleasurable corrigible malfunction".          cowan@x
        --Cordwainer Smith, Norstrilia

Scheme-reports mailing list