Bring out your namespace documents

Volume 9, Issue 123; 18 Dec 2006; last modified 08 Oct 2010


I continue (somewhat sporadically, I admit) to try to get a finding about the TAG issue “namespaceDocument-8” published.

Last time the TAG visited this issue, I thought we devised a relatively straightfoward model of the information that is represented by a RDDL document. There were some t's to be crossed and some i's to be dotted, but it looked like we were approaching closure.

The danger with leaving something untouched for months is that when you come back to it, you basically have to start from scratch. When we did, at the TAG face-to-face meeting last week, I was startled to find adamant objection to the simple model we'd earlier discussed.

Exploration revealed something much more complex.

I think I can see how that's a better model in some academic sense, but I'm not sure it represents a practical way forward for the finding. The finding just needs to ground a couple of alternate syntaxes in some unifying model that's simple enough, and straightforward enough, to be seen as obviously a way of modeling the information in the document.

Not sure what else to do, I took an action to see what's actually out there in the wild as far as RDDL use. In particular, I'm curious about what syntaxes are actually used and what natures and purposes people really employ.

Pointers, please.


I'm not sure whether it counts as "the wild", but there are several namespaces defined by W3C specs which use RDDL documents at their namespace locations. (You knew that already, of course.) On the whole, I think that's a Good Thing.

That said, I can't help regretting that RDDL abuses XLink semantics so badly. It uses the simple link notation, which means the links claim that a particular relation exists between the resource they are pointing at and the contents of the link element. That seems clearly false: the only reason one has to be interested in the links in a RDDL document is that they express a relation between the resource they point at and the namespace whose name is the URI of the RDDL document. As specified, RDDL is a set of rules for systematically lying to XLink processors. When I have pointed this out to people who say they are deeply interested in semantics, they say "Oh, well, it's close enough." Which makes me very skeptical about the future of machine-processable semantics on the Web.

A second problem I have with RDDL is that it seems not to be able to say very coherently what it means by a 'nature' or a 'purpose'. Because of that, when I have to prepare a RDDL document I look for a model I can follow very closely, and I mostly limit myself to pointing at schemas.

But it's still the best alternative I've seen. As in: the only. So good luck to you in your effort to make it semantically sound.

—Posted by C. M. Sperberg-McQueen on 19 Dec 2006 @ 01:53 UTC #

Thanks, Michael. I know about some of the W3C namespaces. I've also heard that Microsoft is using RDDL now. I expect I can find more with a search engine. I'm hoping this essay gets me pointers not just to extant RDDL namespace documents but ones that are actually used by people and software.

Regarding your concern about the abuse of XLink, I have mixed feelings. On the one hand, your analysis is entirely sound. On the other, the use of URIs for role and arcrole clearly make the machine understandable meaning of the link clear. On the whole, an alternate syntax that didn't use XLink would probably be semantically cleaner.

As far as your concern about the meaning of nature and purpose, that might be more clear in the more complex model. Or not. When the TAG minutes are published, I'll add a pointer to them.

—Posted by Norman Walsh on 19 Dec 2006 @ 11:45 UTC #