Johannes starts off with a discussion of REST because that's critical to his design principles for LID. He describes it this way: "everything that matters on the Internet has a URL, can be bookmarked, can be found via Google, can be hyperlinked, can be tagged, and can be accessed with a browser." People got very argumentative here. REST discussions do that.
Johannes' conclusion: people need URLs. Similar argument to Drummond and XRIs, but with a different conclusion. Johannes gives a use case based on Doc's Company Relationship Management scenario and me trying to find a hotel in Berkeley. This isn't so different than Priceline, but it would be decentralized and user-controlled.
To do this you have to
- I publish a need
- Vendor finds the need
- Vendor decides on an offer
- Vendor communicate an offer
If the need is published as a URL, then Google can be used to find it. If a person has a URL, that is the perfect place to point to other information, including needs that I have. In fact, I did that in a non-structured way when I published my need on my blog.
LID allows you to create as many "identifier" URLs as you like. These can be kept separate or they can delegate back to a canonical LID URL. Traversals to specific data within the URL is specified using XPATH queries on the identifier URL. Format of the return data can also be specified (HTML or XML).
The crowd was untypically hostile on this talk. I think the issue is that people are expecting every proposal to solve every problem when I don't think Johannes is saying he developed the complete package. He's taking an iterative, solve a few problems at a time approach. I'm actually fairly impressed with how this has matured over the years. Even so, the feedback was valuable and Johannes is good at taking feedback, so I think it was a positive experience.