XML on MetaFilter?

September 28, 2001 9:13 PM

an xml threaded metafilter with structure based on the dewey decimal system. whenever you make a post or comment it fits it into the big picture. your xsl transformation(s) output based on criteria chosen like dates of submissions/comments, terminology, subject, or other criteria the client can select using a gui/web form, or their own custom xsl stylesheet.
posted by greyscale to feature requests at 9:13 PM (13 comments total)

and the ability to remove or edit posts/comments you "own" (made) [within some limits?]
posted by greyscale at 9:16 PM on September 28, 2001

and a pony?

If anyone wants to attempt something so ambitious, I'd certainly be open for it. I already have a basic xml feed at xml.metafilter.com, and if you can pull that info down, you can starting building something on a remote server (that's how andre does memepusher, the metafilter AIM bot).

I'm busy working on basic bugs and functionality, leaving little time for something like this.
posted by mathowie at 9:53 PM on September 28, 2001

I'm not proficient in perl enough yet to take this on, but perhaps next year. Need to learn more perl, build something to read in the data and check it, set the timestamp attributes, then test and setup the interface for output. But I doubt it could ever perform well enough to be of practical use even if all of the applications were compiled because the data set is so large. It would be a really interesting experiment though. Perhaps I can do it for a masters degree in informatics or something.
posted by greyscale at 11:43 PM on September 28, 2001

As long as we're blue-skying, I'd like to hear more about the idea. Free-range babblin haead.

What do you mean by "an xml threaded" metafilter? Are the XML and the threading connected somehow? or do you mean (a) marked up with XML metadata plus (b) threaded? Are they independent?

So each post would be labelled with some Dewey classification (or, I assume, some other classification--or more than one?), and then the threading would allow me to gather all the posts under some class? What about posts that fall under more than one?

greyscale, have you looked at everything2? I wonder what that plus some appropriate metadata scheme would be like. Topic maps perhaps? Anyone have any thoughts?
posted by rodii at 1:41 PM on September 29, 2001

The 'visual tour' maps generated by the Wiki at c2.com are cool indeed. An example near the Front Door is here. Is this the sort of thing you mean by 'topic maps', rodii?
posted by stavrosthewonderchicken at 10:25 PM on September 29, 2001

Topic maps.
posted by rodii at 8:43 AM on September 30, 2001

Forget it.
posted by stavrosthewonderchicken at 8:34 PM on September 30, 2001

I'm still thinking, and learning. Come back in 6 months and maybe something will come of it. :)
posted by greyscale at 11:20 PM on October 1, 2001

Think out loud so the rest of us can think too.
posted by rodii at 7:43 AM on October 2, 2001

OK:
A co-worker has the meta-meta data (ownership) code developed so that could be tweaked and put in place easily for policing by permission/ownership. I think it is all perl.
Need to figure out how Dewey works and if the scope of Metafilter posts would use that entire tree structure or if it is too broad to be of use (ie would all of MF fit in one branch?).
Then look at the current xml feed and see how to transform that to fit whatever common, existing standard for information exists that is most appropriate to organize the threads.
Figure out how to create the most effective connection (visually or logically) for threads when they jump across the tree structure.
The system needs to retain the ability to just have the linear time-based delivery it now has.
Make sure the time stamps are retained for all comments and threads for time-based queries.
Test and optimize the results of all of the above.
Develop web-based tools for using the system with the most common options preconfigured. Include the ability to request/submit custom transformations of the data for output.
Design GUI for tools.
Test and refine based on feedback/iterate.
Hope that is what you were looking for rodii.
posted by greyscale at 10:41 PM on October 2, 2001

OK, but why Dewey? There are other information idexing schemes, and it's not clear that *any* scheme is going to have the be-all-and-end-all functionality and flexibility something like this would need. That's why I suggested (I know, I know) looking in the topic map/ontology literature.

Librarians! Comment!
posted by rodii at 7:14 AM on October 3, 2001

Dewey just because it has been around a long time so everyone is aware of it and could probably picture/use it.
posted by greyscale at 5:24 PM on October 3, 2001

this idea would make it possible to do things like:


structure output by most active node


see only threads started by users you like


see threads not posted by users you dislike

transform the postings into something that could be run as a script/used for script input to email stuff as mentioned in other feature requests


etc
posted by greyscale at 12:27 PM on October 6, 2001