Thanks for the suggestions. Unfortunately, the RDF was normally to be pasted into the posting that would contain the form element you mention.

I'll look more closely at your XML tag suggestion. Thanks.

A workaround for the 'pasting RDF' problem would be for your form to give a javascript URL to paste into the blog that references a javascript on your site that generates the RDF, just as the googleboxes at http://www.edazzle.net/googleboxjs/googleboxjs.asp do.

On the spider front, you could define some bracketing xml tags to go around a post that contains a link to be considered a blogthread, and have blogs opt in so you don't need to spider the world if you don't want to.

Edited by author 09-06-2002 01:43 AM
Status of effort:

My initial implementation plan was to have users access a web site to register a discussion thread, grab generated RDF from this site, and post it into their posting. This would include RDF/XML decribing the discussion thread, as well as links to ThreadNeedle to generate RDF for use by others wanting to reply to the specific posting.

Well, sounds good but has problems.

First problem is, as has been discussed, RDF embedding into HTML bodies is a problem. The processor will display data contained in elements and the only way to avoid this is to surround the RDF in a script tag or to use such simple RDF that all of the relevant data can be described as attributes of an element. Regardless -- using the RDF will prevent the web page from validating.

Still, this will work, except for a second problem -- weblogging tools want to add white space HTML to all content, and if you past RDF into the posting space, HTML breaks will be added. This will break RDF parsers (the XML is no longer valid RDF at that point). The only way around this is to no have line breaks in the generated code (not workable) or to turn on automatic line breaks (not feasible for most webloggers).

With the release of MovableType's TrackBack, they did highlight another option I hadn't considered (too close to problem, too dense, or both), which is to embed the RDF into the template rather than the posting, and use the weblogging tool's template tags to provide the relevant data (URL, title, author, date). This is great, except for one thing: ThreadNeedle isn't tracking postings, it's tracking dialog threads.

A dialog thread is one person writing a posting about another posting, providing link to same. Sounds simple, but there is nothing that tracks these threads outside Daypop or Blogdex, and they only track one level's worth.

I can embed RDF into the templates and this works great, but there is no way to embed the fact that a posting is related to another posting.

So, I have and am looking at yet more options. Among these is just plain scanning weblogs for links, keeping a record of the URLs and then gradually build the dialog when new postings come in that contain these links. This is how Daypop works, except we would show linkages beyond one level. This approach is very workable, but it will be data and space intensive -- too intensive for my machine and I can't afford another upgrade.

To be honest, I also don't like the idea because it's just plain inelegant. ThreadNeedle is for tracking dialog threads, not postings. It doesn't really care about individual postings. The data store would be full of postings that aren't part of a dialog, taking up space. Even with a garbage scheme, the solution is inelegant.

There's another simple solution, technically, and that is create a form, have people input their posting URL and the URL they're replying to, forget all the folderol and just show the discussion. This is trivial technically, except that we're centralizing the solution, which I wanted to avoid, and people have to add discussion threads. I'm finding that webloggers really don't want to have to manually screw with things. Their interests fall considerably if the process isn't automated.

Another problem with this goes back to AKMA's concern about garbage being introduced. As you can see from this Quicktopic, with a certain teoti, anybody can attach anything. This happened to me with trackback when someone, accidentally I believe, did a trackback to one of my postings from theirs that had nothing to do with what I wrote.

The earlier discussed approach of scanning for links would prevent some of this bogus dialog stranding because you would have to have a physical link in the posting for the dialog.

I thought about piggy backing the technology on to Ben & Mena's Trackback (Movable Type at http://www.movabletype.org) especially with their new stand alone server. But again, there is some complexity associated with this for the webloggers, more than I think many are going to be comfortable with.

Another area of concern with this technology is it does require CGI on the webloggers site, which won't work for Blogspot (and other webloggers without CGI access). We could host a server for those folks, and then working on utilizing the TB pinging. An option.

(I've been chatting with the BlogMD people at http://www.truthlaidbear.com/blogmd/forum/viewforum.php?f=1. They're facing many of these same problems.)

Sorry, I wish I had something I could go "fini!" and throw out to the world, but I've about exhausted all my ideas and technical bag of tricks. So I'm throwing this out to the discussion group to see if anyone can spot something that I'm too tired or too dense to notice. Otherwise, outside of seeing what I can leverage on Ben & Mena's effort (and kudos to them in their generosity in making this technology available freely), I'm fresh out of ideas and ThreadNeedle ends with the RDF schema and no implementation.

Discussion about ThreadNeedle, data, and other issues is also ongoing over at BlogMD at http://www.truthlaidbear.com/blogmd/forum/index.php.

Teoti.com - And you thought you were going outside today -http://www.teoti.com

Yes. Still behind. Sorry.

Hi Shelly, good to hear that your coding as well as your writing continues.
I also got carried away in my initial excitement at seeing this
discussion. However, I remain content in the knowledge that making connections between blog threads explicit should become a reality one way or another. Isn't that sort of what instant outlining was about? Best link I could find: http://radio.weblogs.com/0100887/2002/03/30.html#a157

Needless to say, I'm running behind. Another week or two.

Work is continuing at threadneedle.org. Once the initial prototype is finished - hopefully in about a week - I'll post a note to this discussion group.

The prototype will give us something we can all work with, improve.

Hack and trash, in a nice way ;-)

Server upgraded and work progresses.

Spoke too soon about resources on server. I am going to have problems unless I upgrade. Ouch.

Welcome new discussees.

David, thanks for feedback on queries per /m95. Service should be able to return RDF based on a specific URL, given that the URL is considered the start of a discussion thread. To start at least (resource issue speaketh again)

Edited by author 07-07-2002 10:45 AM
Hi Folks,

I just recently found about about Shelley's ThreadNeedle project. It sounds great!

At essentially the same time as this QuickTopic's been happening, I was independently coming up with ideas about Weblogs and cross site conversations, many of them very similar to the ones mentioned here (darn, I thought I was totally original!). I wrote up my ideas at http://www.dancingrabbit.org/skyhouse/webconversation .

I welcome everyone taking a look at what's there. I'm not attached to the name. I am into the concepts of widely mirrored conversations with signed content.

I'm not sure whether ThreadNeedle as it's currently envisioned could encompass the concepts I'm interested in, but I'd love to collaborate with other people to create a good standard that would be widely applicable.

Edited by author 07-05-2002 04:10 PM
Re /m94:

The service I'd like to see is the ability to ask "What posts reference this post?" This could just be a simple web service that returns a list or links or a bit of RDF in response to a query. For example, http://thread.org?refersTo=http://example.org/blog/20020703%23p4 would be a list of posts which refer to http://example.org/blog/20020703#p4 . (The format of the response could be plain text, with a URI on each line, or RDF or something else.)

Two user-level services that third parties could offer are weblog mediation and thread overviews.

Mediation is the sort of thing Crit does. You would ask it for http://thread.org?page=http://example.org/blog/20020703 and it would go grab a copy of http://example.org/blog/20020703 and after each post it would insert a links to posts which come after it in the thread. (Edit: After creating the modified copy of the page, it would send that to the user.)

For thread overview, you would ask for a page such as http://thread.org?start=http://example.org/blog/20020703%23p4 and it would show the thread following http://example.org/blog/20020703#p4 . (What "the thread" means depends on the algorithm thread.org is using. It could be forked or linear or some sort of graph.)

I have a question for those interested in UI aspects of spec - if you could define the service that ThreadNeedle would provide for querying data from external tools, what kind of service would you see?

(i.e. how would you invoke, and what would you get and in what format?)

Hi Jon, not sure of your question - could you provide a bit more detail?

Alex, interesting concept of voting to strengthen or weaken the thread to a post.

Edited by author 07-04-2002 01:40 AM
Hey, I can't believe that I didn't know about this discussion ealier. I am reading through all the posts, and hopefuly will contribute soon!

I have been approaching the subject from the point of view of having visual forums where each post is a node in a graph. The point here is that users could reply to multiple messages (and those connections would be made explicit by showing edges between the post and the messages that were responded to). Also, one could make a passive change by voting to strenghthen/weaken the tension in any edge (weakened edges would eventually disappear).
  
For more, please take a look at my followup comment to http://nooface.com/articles/02/02/04/1830239.shtml