On Thu, 2009-05-14 at 10:55 +1000, Ramana Kumar wrote: > I think a standard interface for resolvers (i.e. specifying where to > find the library forms that define the libraries you want to import) > would make for a useful SRFI.
I've been working on an SRFI proposal for resolving libraries to files. I submitted it over two months ago and waited one month for the editors to review it and post it as only a draft so it could be publicly reviewed and discussed, but they must have been too busy because they never got back to me, which actually turned out to be for the best, I think, because Aziz pointed out a serious defect with combined version constraints of imports and having multiple versions of a library available. I very much want to support having multiple versions of a library available, so I'm currently stewing about what can and should be done about it, and so I told the SRFI editors to wait for me to get back to them. Other than this problem, I think the rest of it (modulo a few minor things which just need to be touched-up) is ready to go. What I've got so far is at: https://code.launchpad.net/~derick-eddington/+junk/library-files Everything in it about versioning will be changing somehow, maybe totally removed. > We can make a list of possible resolvers, if that would be helpful: > - in a directory "/some/path": look for files > /some/path/library-name.ext with one library in each file. maybe can > choose the extension, and some function on the filename. > - in a file "/some/path/to/file.ext": look for library definitions > within the file > - both of the above could use urls as well as filesystem paths > - in a database > > Would there be any problems with trying to standardize an interface > for this kind of thing? Probably: complexity, corner cases, long-term robustness, usefulness for unforeseen integration with unknown things, making everyone happy, etc. -- : Derick ----------------------------------------------------------------
