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

Re: Plan for revising

This page is part of the web mail archives of SRFI 103 from before July 7th, 2015. The new archives for SRFI 103 contain all messages, not just those from before July 7th, 2015.

On Oct 7, 2009, at 8:51 AM, Derick Eddington wrote:

I'm going to revise this SRFI as follows, unless someone thinks
something should be discussed further.

Remove all design related to versioning.  Add a comment about why
versions are not in file names: it's too controversial.


Require encoding the set of characters: #\%, #\., #\x0 to #\x1F, #\<,
#\>, #\:, #\", #\/, #\\, #\|, #\?, and #\*.  Require not encoding all
other characters.

I think this is fine though the list looks pretty arbitrary.
(for example, why are #\{, #\}, #\[, #\], #\(, and #\) not
encoded?)  I think it's fine that if a library name does not
map to a file name accepted by some who-knows-what file system
or operating system, too bad, pick a different name or curse
the system.  If you have reasons for picking these chars for
encoding and not others, you probably need to say why, in a
footnote or endnote perhaps.

Implicit file name prefix is "main".  Avoid conflicts by mapping
(--- main) to "---/_main.sls", (--- _main) to "---/__main.sls", and so


Ordering of an implicit file name match relative to a non-implicit file
name match is not specified.


Rephrase to state that files conforming to this SRFI must have only one
library per file.  (Thus allowing implementations which support this
SRFI to also support files which do not conform to this SRFI.)


Rephrase to state that cross-implementation files, i.e. those without an implementation-specific file name extension, must contain the library as a standard library form as the first syntactic datum which the standard
read procedure would return.

This sounds fine by me.  I'd like to know Will's opinion.

State that implementations must support implementation-specific files
with the same format as cross-implementation files, but implementations
may also support other formats.

The "must" should be "may".  I can imagine an implementation that does
not wanting to provide implementation-specific files.

(Because of other parts of this SRFI,
there are implied restrictions that implementation-specific files
conforming to this SRFI must have only one library per file and that
such files' path name must represent the name of the contained library
(excluding the version).)

I don't know what you're referring to exactly but I don't think you
should say much (or anything) about what goes in implementation-specific



Remove the restriction that search paths must be independent.  (We
haven't discussed this. Thinking about it more, I don't think there is
enough justification for this restriction.)


I believe these changes will make a much better SRFI.  Thanks for all
the effort.