Re: fundamental design issues (long message) [files] Per Bothner 03 Dec 2005 18:43 UTC

Matthew Flatt wrote:
 > 1. On the issue of storing source, there is firm opposition to mixing
 >    any notion of files into the standard.

I don't think that is possible.  Both R5RS and presumably R6RS have
notion of <program>, which is a sequence of characters that can be
parsed as definitions, expressions, and comments.

Such a sequence of characters is a "file", in the abstract sense.

The real issue is "file names".  On our context, "files name" defines
a partial mapping from "names" to "files" or <programs>.  Some files
have multiple names; some have none (e.g. a terminal).  It is hard to
conceive of a Scheme environment which supports libraries yet doesn't
have files in this abstract sense.

The proposal requires a separate mapping from library names
to <libraries>.  I.e. a duplicate mapping which cannot be
defined in terms of the existing "file name" mapping.

Perhaps there are good reasons for having multiple libraries in the
same file.  But don't pretend you're avoiding "any notion of files".
A library with its associated name is just another kind of file and
the library-name->library mapping is just a different kind of file
name.  You're just requiring the Scheme implementor to define their
own mapping instead of using the whatever file name mapping they're
already using.  Perhaps the extra mapping is useful, but I'm
skeptical.
--
	--Per Bothner
xxxxxx@bothner.com   http://per.bothner.com/