Separation of concerns

There is a great deal of confusion over the who should have responsibility for creating URIs for documents, database records, and the subjects of these "information resources".

We have spelled out what we feel are reasonable requirements for URIs to be used for datainteroperation, and any URIs that meet these requirements should be used. However, this conflicts with the widespread notion that publishers should be responsible for naming the entities they create. When publishers can meet these requirements - clear denotation, access via HTTP, accountability, and open access - we see no problem. But it is rare that a publisher or database provider would be interested in all four of these, and there is no reason that they should be. Publishers usually have little interest in denotation, shift their priorities and go out of business, and are justifiably reluctant to release their materials for mirroring purposes.

While it is desirable for publishers to take on these responsibilities, we do not feel it is reasonable to ask it of them, and for this reason we encourage community-based naming for private (or government) resources when this a publisher cannot embrace any of the four requirements.

One reason this is so confusing is that we confuse a database record that is about a thing with some documentation that tell what a URI names. These are very different beasts, even if the name names the database record. If you want to know what the URI names, you want to know that it names a record published and maintained by some organization that can be accessed via such and such an interface (perhaps one or more URIs, perhaps at one or more mirrors). You want to know whether the URI names some fixed version of the database record or if it names the latest of a regularly updated sequence of record versions.

If you want to know what "Magna Carta" means you go to a dictionary or to Wikipedia. If you want to read the Magna Carta, you find it, or a copy of it. Maybe there is a copy of theMagna Carta in the Wikipedia entry, in which case both functions are served by one document; and perhaps the phrase "Magna Carta" is defined (in its modern sense) in some copy of the Magna Carta, although you would be lucky if that were the case. You can be pretty sure that the original publisher did not define "Magna Carta", although if they had been inclined to the world would be no worse off.

The problem is even more pronounced when the thing named is not a publication or database record. In this case there is no obvious single place to store information about the thing, and in fact different databases may want to store different information about the same thing. Even if one publisher takes up the job of documenting the thing's name (and convinces the community that its names should be used), it will not be acting in the community's interest unless it all sources of information about the thing are treated on a more or less equal footing. The documentation for the name in this case should either aggregate known information sources or supply a bibliography - and would have to include leading viewers to competitors' publications.

One effort at separating concerns is the Common Naming Project.