It's Like Déjà Vu All Over Again
"You could probably waste an entire day on the preceding links alone. But why take chances? We also give you Paul Snively..." — John Wiseman, lemonodor
The "Tivo for Weblogs" idea is very interesting and the phrase itself is an example of what is needed. Google doesn't handle the preposition "for" very well, returning weblogs about Tivo. Just a dash of more sophisticated textual analysis would go a long way, particularly in the smaller scale communities of like minded bloggers.
The mission being discussed in this blogathon is a bit tricky to nail down... For recreating as much, as quickly, with as much impact as possible of the Xanadu ideas, a fully customizable, world class browser would win hands down over something lacking the ability to make use of the existing web content.
The area where I see the most hypertextual potential is in the availability of cross-site scripting in Mozilla. Going beyond referrers into mining your web travels seems very much akin to the "Beyond Backlinks" Ruby note and a handy way to evaluate a blog's interest via an aggregator.
Your comments about TclKit are appreciated... I did a bit more digging, and I see the appeal. On the Mozilla side, XPCOM does allow scripting in other languages but this requires a custom build.
The persistence layer is clearly less abstracted than MetaKit, but I expect this to develop. And for rapid prototyping, the XUL persist="attribute1, attribute2" and simple in memory RDF flush to file works. For mail and history, Mozilla has to use more robust file level storage systems that are clunkier.
And finally, Javascript is a whole new ball game in Mozilla. With the debugger and DOM inspector coding JS is much nicer. I've got a persistence, gui, and scripting Mozilla FAQ in the works.
For the aggregator scenario in Ruby's note, authoring is a key component... Mozilla.org has yet to really discover the killer app for Composer other than 4XP (Communicator 4.0 parity) though we're getting there. [YACCS Comments for It's Like Déjà Vu All Over Again]
Ah, I think I understand better what the issues are now. It seems we're conflating two different project ideas: Sam Ruby's collaborative filtering for weblogs idea, and my Xanadu-prototyping idea. I definitely see them as quite distinct: Sam's very definitely revolves around the web and existing standards and protocols (HTTP, HTML, XML, RSS...). I think this would indeed be an excellent project for Mozilla, as it seems like the obvious path to wish to be able to take would be to extend the web browser to do what Sam suggests.
I've already commented fairly extensively on the Xanadu-prototyping idea, but one thing that I didn't really discuss much was the relationship between networking and persistence: it occurs to me that the reliability of a system such as Spread goes nicely with the transactional support of TclKit. I believe that would be a crucial combination in supporting the various data structures and algorithms from Palimpsest or Xanadu or whatever we wound up prototyping.
The critical point with respect to the Xanadu prototyping idea is that such an endeavor already presumes that we've gone too far down the path of worse-is-better and W3C TLA's, so it's necessary to pull back and attempt to begin from a very different set of fundamentals. At the time that Xanadu was originally launched, issues like process group communication and persistent storage were themselves utterly open research topics. They're still research topics, but at least now we have some research systems to build on instead of having to develop even that level of infrastructure ourselves.
What I'm suggesting, though, is that unlike Sam's idea, which is definitely an application-level idea, we undertake something much closer to pure R&D, but in a very narrow context: just Xanadu and/or Palimpsest data structures and algorithms. When it comes to networking, persistence, GUI... we use the smallest, simplest, most loosely-coupled components that already exist that we can find. It's these criteria that lead me to propose tools such as TclKit and Spread and make me leery of Mozilla in this particular context.
6:44:54 PM
