> On Oct 25, 2022, at 2:10 PM, Arthur A. Gleckler <xxxxxx@speechcode.com> wrote: > > On Tue, Oct 25, 2022 at 10:26 AM Marc Feeley - feeley at iro.umontreal.ca (via srfi-discuss list) <xxxxxx@srfi.schemers.org> wrote: > > To me the terseness and elegance of the notation is important so the choice between these two is clear to my eye: > > (import (github.com/gambit/hello demo)) > > (import ((uri "https://github.com/gambit/hello") demo)) > > Moreover, currently the first notation is allowed by R7RS for a library name, but not the second, so nothing would have to change in R7RS. We just need a convention for the community to adopt. For example you can use (github.com/gambit/hello demo) as a library name in Chibi Scheme and it will look for the file github.com/gambit/hello/demo.sld in the module search directories. This means that an independent installer could be written for Chibi to dowload the sources and put them in one of the module search directories (without having to change Chibi). This is probably a few lines of shell script calling wget/tar/etc. In principle that notation should be supported by other R7RS conformant systems although the presence of a “/” in the name may trigger bugs in the implementation (the R7RS says it is “inadvisable” but not “an error” to use “/” in a library name). > > Yes, but what you're proposing changes the first element of library names from a simple identifiers into something structured that has a special interpretation. That's why I prefer the idea of using an extension of the standard library-name syntax. Relying on the fact that many URIs are legal identifiers feels like a hack. On the contrary! It shows the foresight the RnRS authors had in allowing so many characters in symbols. Moreover, note that the delimited symbol syntax |...| allows any character to be put in a symbol (just like the "..." syntax for strings), so a symbol is just a more “human friendly” kind of string and it works quite well for URIs (I have never encountered a web site where a delimited symbol is needed, probably because it is in the interest of web site creators to use a name that is human friendly). > > Note that this approach to naming libraries is also used by GoLang, so there’s something to be said about its practicality. See https://encore.dev/guide/go.mod > > Ha! Citing a project of the AT&T crowd as an example of good syntactic style will never convince me! And yet you use Unix (I assume)! > Would someone else like to join us in this bike shed? I’m sorry to hear you feel discussion on the library naming issue is bike sheding... I think it is a critical issue that is hindering the sharing of Scheme libraries. Having a standard layout for publishing libraries is fundamental. Without it the total work required is O(N x M) rather than O(N x 1), where N is the number of Scheme implementations and M is the number of library layouts in use (and it seems M is close to N). Marc
