Magnus Manske wrote:
OK, if you're all in favor of that, I might be
willing to code (what a
surprise!). I deeply regret, however, that I can't remember the
details of Erik's scheme.
Before you go looking for the original mail, was it something like this:
* A [[category:foo]] *link* in an article works similar to a language
link (shows somewhere in the surrounding framework rather than in the
* The [[category:foo]] *article* has a description of the category,
and some kind of automated list of backlinks.
If Erik suggested something completely different, please find that
mail for me...
ABOUT THE CODE
That sounds pretty much like the Erik's proposal as I remember it, and
it seems both harmless and non-intrusive to existing work. Yay! Let's
There was at one point a proposal for a separate DB field for metadata,
like the language and category links. It's a nice idea for the long
term, but it's simply too much complexity for now, and would complicate
things too much by requiring lots of new UI code. So let's just keep the
category links in the main body, and keep them at the top or bottom by
* The description of the category should be in the "page text" for
[[Category:xxx]], like some of the special pages.
* category page bodies should be able to be edited by anyone, just like
ordinary pages, with the same code for protection by admins, deletion, etc.
* category page bodies should also be able to contain category links
themselves, allowing a directed graph of categories to be created.
* there should be a "list of all categories" special page
* moving category pages does not really make sense in this simple scheme
as the names are "hardwired" in the [[Category:XYZ]] links in the
* but there's probably no reason to disable this operation...
Once we have some code, the next thing to do is to agree how to use this
feature: but this should not hold up writing and releasing the code.
* lots of arguments about category schemes
* pragmatic agreement on the most obvious categories
* much useful entry of data according to various schemes
* eventual and wonderful automatic indexes, Bayesian
auto-classification, etc. etc.
I'd like to suggest a few conventions, before the category wars begin.
There are a number of existing category schemes at the moment, such as
the Library of Congress system, and SUMO. I suggest that, rather than
using these codes raw, we prefix them with their scheme name, thus:
[[Category: SUMO ExistingThingName]] or
[[Category: LOC BD-300]]
This will greatly simplify collating them together, or
machine-processing them later, and helps forestall the flamewars about
Dewey vs. LOC vs. Cyc vs. SUMO vs. ... each set of enthusiasts can
simply code away as they like...
This leaves the "no-prefix" namespace ready for the Wikipedia category
My first suggestions:
[[Category:Person]] eg [[Henry Droop]]
[[Category:Place]] eg [[Poland]] or [[London, England]]
[[Category:Time period]] eg [] or [[15th century]]
...and then we can use the information from the more sensible [[List of
XXX topics]] articles, for mathematics, physics, politics, medicine,
rare diseases, etc. to auto-mark the given articles.
If we do this right, we can probably _eventually_ get rid of most of the
list articles entirely: ''but'' there is one fly in the ointment with
doing things this way: we lose the ability to create links in category
lists to non-existing articles. This would be a great pity, as it is one
of the most useful ways to create lists of new candidates for articles
within a topic.