[r6rs-discuss] library versions (was: Rationale issues)

From: William D Clinger <will>
Date: Fri, 29 Jun 2007 06:29:47 -0400

Lynn Winebarger wrote:
> Is it just me, or does the sentence (from 7.1 of R5.96):
>
> When more than one library is identified by a library reference, the choice
> of libraries is determined in some implementation-dependent manner.
>
> seem useless? One might even think it implies the choice of libraries might
> ever not be determined in some implementation-dependent manner.

I don't think it is useless to note that it is impossible
for a programmer to write a library reference that cannot
identify more than one library. From that it follows that
a programmer cannot avoid reliance on some implementation-
dependent mechanism for mapping library names/versions to
concrete libraries. From that it follows that there is no
such thing as a truly portable R6RS program that uses any
libraries beyond those described in the two potentially
normative reports.

Spanky will require all non-standard libraries to be part
of the same file that includes the top-level program.
Other R6RS-conforming systems may not allow this. Hence
programs that rely on non-standard libraries are not
portable.

This isn't fatal, because the proof above establishes the
non-portability of program representations. There remains
the possibility of a conceptual level on which programs
might be portable in concept.

Will
Received on Fri Jun 29 2007 - 06:29:47 UTC