Mailing List Bots
Ben Hammersley, ORA, etc

Feature creeped away from mailing list bots uniquely. Message Threading in general now.

Interested in structured data in general, beyond his work on RSS book for ORA. In looking into RDF/SemWeb over last few years, determined that it started in the wrong place. The web is the worst place to start looking to structure data.

So where is data already structured for semantic meaning. The obvious place is mailing lists. Which include "lots of yummy metadata" in headers. Explicit and implicit Information.

Ex: Yahoo Groups mailing list header. Includes self-identifying info, actual location, yahoo id, locations, etc.

Citation of Doctorow Metadata as Metacrap. In unstructured web, this is true - never going back to correct information done in the past - lucky if we do that going forward. But in email, this is lazily included, or can be easily inferred.

Further, mailing lists have further advantages. They are chunky conversations. As is IRC and weblogs. In IRC, smaller messages. Get date/id/subject (by room -- I question this contention). Weblogs have heaps of metadata as well, including links, etc etc.

RSS feeds - RSS 1.0 in example.

digression - history of RSS

Contains much good info, obviously. And RSS can be used to map mailing lists like data. Discussion of what is captured, including commentors, trackback, referents, full content, etc. So you can take a flat document with metadata that in turn points to other flat documents that are (directly) related to it.

RSS is extensible. Ben is showing RSS that contains contributor information (using <A HREF="http://www.windley.com/2003/04/23.html#a574">Friend of a Friend</A> data), trackback data, and annotation references. All of this is added automatically by the system.

Can take similar technology and use it to represent message Threads - ThreadsML. RSS1.0 with extra toys. By utilisation of additional categories - mod_threading and dublin core.

(dublin core as provider of namespace for core information)

Mod Threading (link: http://f3.grp.yahoofs.com/v1/QEKoPvY589C_jgRhsxc36lITVFzs3r3XsEx66p-rpCnMp5FTLpulCkyEwORp0t51-ZqEjHu1m6mknR8X/Modules/Proposed/mod_threading.html) See also http://www.benhammersley.com/archives/002981.html

    new idea that allows pointing to children in a thread. (aside - is reference only to children a weakness? Arguably simplest to architecturally represent).
    

    Children do not imply ordering.

A representation of the ongoing conversation. Can be thrown into a large dataset, or refer across items. Interoperable standard for any chunky data conversation.

Those places that become most trusted are conversational - discussion groups, mailing lists, etc. If these trusted threads could move between systems, a joined semantic web WITH PURPOSE could be built.

The major problem is subjectivity. Subjects are rarely useful, individuals will provide their own unique categorization.

One answer is heirarchies (dmoz, yahoo, etc). But this doesn't work going up. They are culture specific. One man's weblogging is another person's webbity. This creates heirachies as brittle tools.

    I've posted the notes from a morning discussion today involving Ben and a few others, on the topic of ontology, at http://www.crystalflame.net/archives/000052.html -- it's rather interesting, associated with the topic.

A possible answer - ENT/RSS2.0 - could be used for creation of a heirarchy based on what you know, and sharing it. Released last week. (there's a link on molly.local wiki if someone wants to dig it up)

ENT posits a cloud reference for the ontology. (akin to a XTM). Provides classifications under the terms of what you wrote about. Removes the cultural specificity by bolting the ref into the item in play.

KCollector -- http://www.evectors.com/ this url is close, need to ask benham for more specific --rsoderberg
    RSS aggregator for RSS feeds containing ENT data -- "topic clouds"
    Five sites using it, 50 keywords so far

http://matt.blogs.it/specs/ENT/1.0/
http://www.purl.org/NET/ENT/1.0

Could lead to so many ontologies and categories that the diminishing utility is rapidly reached. Open for discussion.
    Useful ontology reference sites would gain popularity, like dmoz and yahoo. --rsoderberg

But so many past attempts at ontology standards have failed! Audience member doubts about whether ENT is The Answer...
    See the crystalflame notes (linked below) from this morning's discussion on this as well; it, in a rather scattered form, addresses why it might work for this system: "metamantic web", let the software build the ontology relationships -- since humans can't do it, as past attempts have shown. Can computers? --rsoderberg

Other emergent taxonomies.

  "More Like This From Others"
    @ the MovableType plugin registry: http://www.mt-plugins.org/
    By dint of linkage, builds ontology; passive reputation for each ontology is built.
    q: how is this different than anchor text lookups?
    a: the link is between one post and another, but the reputation broadens to the realm of the site - two clouds, not two points on a line.
    

  Other sources
    FOAF - spec: http://rdfweb.org/foaf/
        RDF descriptor for people and their attributes - their interests, location, OS, etc.
        see also: http://www.sixapart.com/log/2003/01/fun_with_foaf.shtml
    

    GeoURL - spec is at http://geourl.org/
        similar, for geographical location
        RSS, RDF, XML, http meta tags
        Has an RDF-output RESTful interface, using the FOAF schema! @ http://geourl.org/news/archives/000025.html
        see also: http://geourl.org/news/archives/000023.html
    

    Ratings
    

      Showing of agreement or disagreement
      undefined as of yet, much discussion.

    ThreadsML

      Stalled in November, Marc picked up on it hoping for a critical mass of people at ETCon.

Marc Cantor: What are the human-oriented benefits to this?

  TopicExchange could be used well for tracking of the ideas.
    Phil Pearson has an open server based out in New Zealand called "The Matrix"

  Weinberger: Easy conversion of an email trhead into a threadsml document space, viewable in whatever system you prefer. Allow for ready reentry.
  Adwords as economically viable way to sponsor
  Same conversation happening in world of networks (domain names, for example?)
  Two ideas from one guy
    one: use IM client to respond to an email
    two: show up on a site, just be able to subscribe to the thread
  Lazyweb idea: filter a blogroll based on topic dynamically.
  Do these threadings on a local system or set of systems, rather than on open network. (ben references AgentFrank, RDF local repo, tracking into database)
  Marc C: this should work for graphics and music
  ??: wants to be able to associate a set of documents into an email, and should be able to automatically associate the backstory between interactions.
  Pete: Use thje