Google, directories, OPML

This was a ramble that started on Scripting News, part of the ongoing conversation about the Google API, and ways of connecting content management to Google through SOAP. I moved this to SoapWare.Org. I'm going to edit the ramble and see what happens.

Feature tour 

Boy one thing the Google API did was get me thinking about more ways to integrate content management with what they do. Key point. Google is more than a search engine. It's also a directory and Usenet and news (and images but I've already written about that).

Focus on the directory. I love directories. Remember I'm the outliner guy. OK, so how to make the directories work. Distribute them the way you distribute the Web. One simple trick is needed. When Google finds an OPML file, it should branch into a directory displayer like this one. I don't hand-code directories, I use an outliner. Click on the screen shot to the right to see what it looks like on the editorial side.

No one has cloned this, as far as I know -- all the OPML displayers so far view outlines as documents. A mind bomb awaits you. Yes, they are documents, but viewed another way, an outline is a mini-website. This directory for the site you're reading right now is a single outline. Look at this screen shot to see how it's edited. See what's going on?? The whole site is a single outline. Easiest site to edit ever. We do lots of them. That's a key feature of Manila that Blogger, Slash and Movable Type, etc don't have.

All the directories in Manila are available in OPML and can be opened and browsed in any OPML-compatible outliner. See the white-on-orange XML button on the directory? It links to the OPML version of the directory. Maybe it's time for this rocket to blast off.

Why this is the Googlish way to do directories 

First, I think Google was right to latch onto DMOZ to get their directory effort bootstrapped.

But the problem with DMOZ is that it centralizes control of the directory. But there could be dozens of ways of viewing any given topic, and the Web (and Google's search engine) use page rank to give authority, and that's dynamic and valuable, but the DMOZ way freezes authority in time, and doesn't let the collective human intellect of the Web determine authority as the search engine does so well.

The power of the Web that Google's search engine taps so well is that it is totally decentralized. The same can and should and must happen with directories. We scatter them around the Internet, each CMS learns how to browse them, people choose which ones they want to contribute to, and if the editor is responsive and gives people what they need (without giving them too much), it rises through the ranks.

Anyway the more I think about Google, the more I think we're just inches away from a much more powerful Web if we can connect what we have working with what they have.

My theory of the Web 

There are three useful ways to view the Web: 1. Search. 2. Hierarchy. 3. Time.

Google is the best at Search. It's weak at hierarchy (like everyone else). And it has nothing for time. (Weblog archives and Archive.Org are the two best approaches to time so far.)

If we approach hierarchy at the microcontent level, instead of trying to create uber-directories, which clearly don't work, we'll nail problem #2. Net-net, if when Google encounters an OPML file it's ready to browse it as a directory, we blast through #2.

A short-step on that path is for other developers to explore generalized HTML-based directory browsers as we have at UserLand.

Directories and inclusion 

Now for the one tenth of one percent of the people who still understand what I'm talking about, outline directories are as spacy as or even spacier than the Web itself. They have the concept of a hyperlink, but a hyperlink is much more powerful in outlines because there's no need for a visible seam. When I click on a link in an HTML browser the whole context shifts. When I expand a node in an outline, it just fills in underneath the link. The same concept applies to HTML outline browsers. The Implementations sub-directory on this site is actually edited by Paul Kulchenko. But unless you look carefully, in every way it appears to be part of this directory.

Even deep-linking is possible, and as with the HTML web any OPML document can link to another, creating the opportunity for many uber-directories, containing all the content they want from other directories. If you want to have a SOAP sub-directory for what you do, just hyperlink to this one, and voila, you're done.