Posting to RDA-L
Karen Coyle wrote:
One hint, though, if I may, is that the goal of linked data is NOT to then put the data in a database. The goal is this one that you list as the third rule
> The third rule, that one should serve information on the web against a URI …
is the goal: to make your data available on the web. That means not in a closed database, but actually on the web. It’s like putting your documents on the web so that anyone can link to them, but in this case you are putting your data on the web. Because each “thing” in your data has a URL, the web allows you to make links.
But it is my understanding that you can go through a database to get to the data, and as a result a URI includes the OpenURL. This is a “relative reference URI”, where you have to establish a base URI. See http://tools.ietf.org/html/rfc3986#section-5.1.
Here’s an example OpenURL I found, which I think everybody on this list can understand:
The big point is the “?” which means that this is a query, or a search in a database. OpenURL says that everything *after* the “?” should be standardized (i.e. any database should be able to accept this type of standardized query) while everything before the “?” (i.e. the base URI) can change.
This is why I have thought that OpenURL demonstrates the power of consistent, standardized metadata: so long as everything is consistent, the data in one database can be used to automatically find and reliably query another database. But of course, if everybody follows their own rules, the entire thing breaks down. In the example above, for the volume number, is it “24” or “XXIV”? Notice also the author’s name, which follows UNIMARC practice of coding first and last names separately.
When the base URI changes, as they always do, so long as everything is set up correctly, all you have to do is change the part before the “?” one time in your wonderful relational database, and all is fine since the rest is created automatically from the record itself. And since it is standardized, you can search any and all (in theory) databases just by adding new base URIs. In this way all you need is highly standardized data, which is the specialty of catalogers.
I have always thought that OpenURLs should be very powerful tools for catalogs, and it is crystal clear to everyone that if you use “Mark Twain” and the others use “Samuel Clemens” the system simply won’t work, so the power of consistent standardized metadata becomes desirable even to the public.
Concerning linked data, it is simply the way the World Wide Web works, plus there is the assumption that the more links a resource has both to it and from it, the greater use it has for everyone. I am sure that almost any page you see on the web is composed of dozens of separate files brought together from all kinds of places: within the site itself with such files as headers and footers, images, and from other sites on the web. Some of these files in turn import other files from other places, which can also import still other files, and on and on.
Linked data actually does more or less the same thing, so it is nothing really new from the technical standpoint, but it is based more on semantics and meaning. So, if somebody wants pictures of Rome, the system should be “smart enough” to know the different forms to get you Rome, Roma, Rzym, and so on. Of course, in practice this means that *computers* need to understand that these are the same. To make your resources work on with linked data, you do things in RDF, which is a type of XML. This is no different from using style sheets on your website, and you do thinks in CSS.
But all of this is being done now in dbpedia. Look at http://dbpedia.org/page/Benjamin_Spock, scroll to the bottom and you can see the record in different types of RDF.
This is one of the reasons why I think switching to MARXML would be one single step in the right direction, and also why we should really consider working with dbpedia: a lot of the technical work has been done already.