[ Sent to locator-sig, which seems dead, and starship-crew, which is
definitely alive.]

Here's a straw-man proposal for a Python module locator; the HTML
looks rather ugly in Netscape, but it'll give you the idea, and I
simply want to spark some discussion about this; where should it go
from here? Go to:

http://starship.skyport.net/crew/amk/locator/view.cgi

That gets you to the root of the locator tree. There are two
branches, "Data Structures" and "Encoding".
FUTURE: How should the tree be organized? Like the
contrib directory on ftp.python.org, or some other scheme?

Follow the "Data Structures" link, and you'll see a single
branch ("Mathematical"--for sparse matrices or whatever), and 4
packages listed.
FUTURE: Packages currently have 4 attributes: title, version,
description, primary_site. What attributes should packages have?
(See the Linux Software Map for ideas.) A more attractive
rendering--making the primary site a live link to the site, for
example--isn't hard, and will eventually happen.

To maintain the tree, go to:
http://starship.skyport.net/crew/amk/locator/login.html

Use the login 'amk', and password 'pwd1'. If you get them
wrong, you'll get no feedback, and will just see the root of the
locator tree as before.
FUTURE: Perhaps we shouldn't have password-protected
maintainers; perhaps people should submit entries via
Web or email which get approved and processed.

Once it succeeds, all the packages will have DELETE and EDIT
links next to them. There are 4 fields at the bottom which can be
filled out and submitted to add a new package; there's also a field
that creates a new subcategory or branch, at your current location.
So, try it! Create a new package and delete it; create a new category
or subcategory of an existing category. (Don't bother to add packages
for real.)

Wandering around
http://starship.skyport.net/crew/amk/locator/tree/
will give you an idea of how the data's stored; it's just using the
filesystem, and pickling packages into files with generated names. No
big deal...
FUTURE: Perhaps a real database is required? But I
expect we could automatically generate static
index.html pages for each directory, once CPU load
gets to be a problem. Only maintainers need to see a
dynamically generated page (though one could imagine
users being able to customize their view; excluding
packages that don't support the Mac or whatever).

What do you think, Sirs?


Andrew Kuchling
amk@...
http://people.magnet.com/%7Eamk/

_______________
LOCATOR-SIG: Discussions about a Python Locator for resource discovery

send messages to: locator-sig@...
administrivia to: locator-sig-request@...
_______________