2003-04-23 Ben Hammersley Mailing List Bots, kinda Been looking into ways of structuring data. Semantic Web started off in the wrong place - the "web" is the worst place for structured data... nested tables, html3; unparseable. Too big a problem. Where is there that already has structured data? Mailing lists. Mailing list mail headers. Shows headers from a Yahoo! Groups message; metadata. The sender's email, full name, yahoo id, internationalisation, etc. Cory calls metadata "metacrap" - no one's going to enter it all. Too much work, too pointless. When you send an email it's made for you. Mailing list conversations are in chunks. So are those in IRC (date, id, subject (room)). So are weblogs - they're mailing lists that you "write to yourself" (identity, date, etc.) I'm not convinced... Brief introduction to RSS... Netscape syndication, jumps to RSS 1.0. Shows one of his weblog entries in RSS. tells you where to go to disucss the item. , who's linked to the entry. ThreadsML. - mod_threading and dublin core. Pointing at more XML on the screen, some kind of post (email? weblog?). element shows the child posts of this post. Whatever two-way or one-to-many conversation you're having, it can be marked up in ThreadsML and things can be linked together. Obsolescence of conversations: Much of the signal:noise stuff on the web are moderated forums or mailing lists. Problem: Subjects. Searching by topic/subject is difficult. "Subject:" lines aren't useful. Tree-based heirarchies are an old solution, works for going in one direction, but not the other (not sure what he means exactly). Heirarchies are culture-specific and are thus brittle. An alternative is to create your own heirarchical ontology. ENT, Easy News Topics for RSS 2.0. Shows an example item, perhaps referring to an essay about baseball, referencing a custom ontology of the players/managers. It says "This is how I see baseball, this fits into my view of the world." Good ontologies will be picked up by other people, bad ones will fall aside. New RSS aggregator for ENT 1.0 data launched this morning. topicexchange.com, "evectors k-collector beta": http://k-collector.evectors.it/itentdirectory/home?dir=3 Problem: You end up with too many subject headings. Emergent Taxonomies. Some stuff about Trackback. I'm lost. Something about linking posts together. The slide says: More like this from others * I post something within my own category * Someone links to me * I pick up the trackback * Follow the link back to the linking-post * I find the linker's subject * The connection is made When you start merging "lumps" the category name is irrelevant, and so it should be - it's the problem. Sniffing RDF Glue - Additional Sources - FOAF, GeoURL, Ratings.