This is the mail archive of the
guile@sources.redhat.com
mailing list for the Guile project.
Re: Redistributing Guile contributions
- To: djurfeldt at nada dot kth dot se
- Subject: Re: Redistributing Guile contributions
- From: thi <ttn at revel dot glug dot org>
- Date: Tue, 15 Aug 2000 01:12:24 -0700
- CC: guile at sourceware dot cygnus dot com
- References: <E13ORXa-0002AE-00@mdj.nada.kth.se>
- Reply-to: ttn at glug dot org
From: Mikael Djurfeldt <mdj@mdj.nada.kth.se>
Date: Mon, 14 Aug 2000 23:13:02 +0200
Last night I realized that we are extremely close to having a
fantastically simple redistribution system for Guile contributions.
[insomnia -> GUMM design ideas]
*We already have a process which can support such a repository.*
[We can tweak the projects list.]
The only thing which is missing before it can be put to serious use
as the Guile library is a hierarchical structure.
i would qualify that as "actual use". serious use requires validation
of the entries (whether simple checksumming or complicated interface
versioning) so that a library user can have confidence in the library.
(insert explanation of proper librarian duties here.)
on the other hand, actual use is a nice thing, too!
The simplest solution would be to let people add new subdirectories
as they wish. (We can always fix it afterwards.) One field could
simply be a path among webpages. If the components of that path
doesn't already exist, it is created. Note that I'm not talking
about real directories, but only relationships between pages.
being the janitor in a parade is not so fun. in practice, people will
create bad (locally convenient but poorly thought-out) hierarchy and
come to rely on it, then complain bitterly when some poor maintainer
screws up enough courage to reset the bones.
Just to be clear: This repository would not contain any files. It
would only contain pointers out into the net. Later, we could add a
mechanism for checking links and removing them if they have died.
Can you make the necessary adjustments?
although i'm a big fan of indirection, i think a library has to be able
to instantiate its advertized goods. certainly one copy of an entry can
be kept on the server and links be maintained to a second source. i'll
keep your suggestions in mind.
obGUMM: still doing the guile-projects -> postgres transition. reading
"sam's teach yourself sql in 10min", too. today some cager on the SJO
Terminal A loop helped me find more free time by bumping my SO and me
off the motorbike, turning its wheels skyward and generally making it
shop-worthy... ahhh, vacation silly valley style.
thi