Here is our current understanding of what is required of a URI that is to be use in a Semantic Web for scholarship. These should be observed by everyone writing RDF - both users of existing URIs (who are considering what's already available) and those coining new URIs.
- The intended referent of the URI must be made clear through documentation.
- Provision of such documentation via a widely deployed network protocol (see URI documentation protocol) must be an ongoing concern.
- The documentation provider must be responsive to community needs, such as the need to have mistakes fixed and the need for stability of reference.
- Documentation must be open.
1. URI documentation must say what the URI is supposed to name. If the referent is not clear, then statements made using the URI will not be falsifiable and communication will be difficult. (TBD: cite Confucius) If there's no way (at least in principle) to judge the truth of a statement made in RDF, then it is not of much use for scholarship.
2. Documentation should be accessible via HTTP because users will encounter URIs and will need to know what they mean. Users cannot in general be "protected" from seeing URIs, and they cannot be expected to know to use special software or a special web page to find documentation.
3. We all wish that documentation would be correct and complete from the time it's first published, but that will never be true. In addition, it improves the efficiency of science to augment documentation over time with examples, nonauthoritative interpretations, and bibliography. The documentation should be more like a wiki (moderated, perhaps) than a database or textbook - although a textbook in the sense of regularly updated editions is a very similar idea.
Unless explicitly provided for in an "under construction" disclaimer, we insist that the documentation not change in such a way as to imply that the name names something different from what it named before. If a publisher is allowed to change the documentation incompatibly, then a communication over time (e.g. via URIs stored in a document or database) will be impossible as a writer's intended meaning may differ from what a future reader may learn from the changed URI documentation.
On the other hand stability of reference is not the same as stability of the referent! The URI may denote something that itself changes (such as a human institution or a document), and that's OK as long as this expectation is made clear. See Stability.
4. If mirrors cannot be established easily then the URI's documentation infrastructure will have a single point of failure. Loss of documentation hurts not just its publisher but also everyone who has ever read or written RDF that uses the URI.
Documentation providers might consider the Science Commons Open Access Data Protocol as a way to permit free use of documentation.