“Every package has its own home page on the web!”

In his latest essay, Cliff Gerrish riffs on my keynote at the Kynetx conference:

There was a moment when he was talking about a meeting in local government where the agenda was managed using a web-based tool. Udell talked about wanting to be able to hyperlink to agenda items, he had a blog post that was relevant to one of the issues under discussion. The idea was that a citizen attending the meeting, in person or virtually, should be able to link those two things together, and that the link should be discoverable by anyone via some kind of search. And while the linking of these two things would be useful in terms of reference, if the link simply pulled Udell’s blog post into the agenda at the relevant spot, that might be even more useful.

The reason this kind of thing probably won’t happen is the local government doesn’t want to be held responsible for things a citizen may choose to attach to their agenda items. A whole raft of legal issues are stirred up by this kind of mixing. However, while the two streams of data can’t be literally mixed, they can be virtually mixed by the user. Udell was looking at this agenda and mixing in his own blog post, creating a mental overlay. A technology like Kynetx allows the presentation of a literal overlay and could provide access to this remix to a whole group of people interested in this kind of interaction with the agenda of the meeting.

The Network provides the kind of environment where two things can be entirely separate and yet completely mixed at the same time. And the mixing together can be located in a personal or group overlay that avoids the issues of liability that the local government was concerned about.

That’s one example of a general pattern and best practice at the core of what we mean when we talk about linked data. In 1997, Andrew Schulman gave a talk entitled The Web is the API in which he meditated on the then-revolutionary UPS tracking application:

A URL can drive a process. Thus, UPS was not only opening up its business practices to its customers, but also publishing an implicit API (hmm, “the UPS software developers kit”?). For example: http://wwwapps.ups.com/tracking/tracking.cgi?tracknum=1Z742E220310270799.

Every package has its own home page on the web!

So should every city council agenda item. So should every ARRA contract. This was obvious in 1997, it’s even more obvious now.

The other day I listened to Tim Berners-Lee’s 2009 TED talk on linked data. He said nothing about RDF, but a whole lot about HTTP. The message boiled down to: “Give everything that matters its own home page on the web.”

That simple idea is easy enough to understand, but entails much that isn’t obvious. It costs nothing to mint new web namespace. We can name trillions of things, and we can declare trillions of relationships among named things. No central authority will govern these names and relationships. There is an infinite supply of unique names, each a needle in a vast haystack. Search engines can easily find any needle in that haystack, and they can easily find related needles too.

Here’s one more non-obvious point: Naming is hard. When we have to stop and think what to name something, we can end up thinking for a long time. In order to get to trillions of named things we’ll need to automate the naming. That’s part of what I was driving at in OData for collaborative sense-making. Given any set of things, the web names for those things, and for the relationships among those things, need to arise organically from the systems we use to create and share them. Information systems should mint usefully granular web namespace. If the right kind of naming is built in, we won’t have to bolt it on later. Things will naturally form relationships with other things. Views of those relationships will emerge from many perspectives, for many purposes.