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

Re: languages

 | Date: Tue, 14 Mar 2006 14:58:28 -0800 (PST)
 | From: bear <bear@xxxxxxxxx>
 | On Tue, 14 Mar 2006, Chris Hanson wrote:
 | >bear wrote:
 | >>
 | >> So to me a scheme URI would be something like
 | >>
 | >> "scheme://R5RS/TinyCLOS/rationals/foo.scm"
 | >>
 | >> and this woud be a key into a small database that mapped
 | >> it to a URL giving a web address where it could be downloaded.
 | >But why not just use an http: URI to do that?  You don't need the
 | >small database at all.  Something like
 | >
 | >	http://schemers.org/R5RS/TinyCLOS/rationals/foo.scm
 | I think that would imply excessive reliance on a single
 | site and server. And this is sometimes a problem.  Aside
 | from the situation becoming such that the entire community
 | would be hosed if a single server should go down, it would
 | impose a significant burden (and bandwidth costs) on the
 | maintainers of that site.

I believe Chris is suggesting multiple sites:


A less idiosyncratic site for SLIB could be
http://savannah.gnu.org/projects/slib/ or

 | > The database could be stored on the server instead of having a
 | > local copy everywhere.

SLIB maintains a local catalog (database) of named modules on the host
running or compiling Scheme.  "(require 'new-catalog)" rebuilds the
catalog; which is also done if the catalog is missing.  This approach
works well and runs quickly.

Going out to the web for every module request is a bad idea.  Programs
which worked yesterday may fail today if a module has been changed.  A
program will fail to run if a server hosting a module it needs is

A library of interest should be downloaded and stored on the local
host, its entries integrated into the catalog on the local machine.
These operations can of course be automated.