 |
 |
Hi Pete, Pete Johnston wrote: > Yes, you're correct that the prefix name is just a convention, and > there's no "hard-wiring" in the DCMI metadata profile. > > In an earlier draft there was something in the profile transform XSLT to > allow for the DC/DCTERMS prefixes to be mapped even if they weren't > "declared" (using rel="schema.DC" etc) but that was removed in the final > version. > > The spec does currently say (in 3.1.2) > > "Also, the characters used for the prefix in a DC-HTML Prefixed Name are > not significant, but communities often adopt a convention on the common > use of a prefix to facilitate human readability." I think that's clear enough. However, statistics done by Philip Taylow (WhatWG) show that few authors declare their prefixes; likely because tools do not care either. This is bad. > But, yes, I think it would have been helpful to include an example that > explicitly illustrated this point for the DCMI-owned namespaces, just to > remove any potential for ambiguity. > > I'd be slightly reluctant to create a new version of the profile - i.e. > issuing http://dublincore.org/documents/2009/06/nn/dc-html/ and asking > people to refer to that profile URI in their X/HTML documents - in order > to do this, because it isn't changing any of the "semantics", rather > just clarifying a point of the existing spec. Yes. The profile URI should be stable, unless incompatible changes are made. (As a matter of fact, I didn't realize that the profile URI *is* the specification URI until now) > What I suggest we might do is to amend the content of the existing > document, keeping the same URI, with an "erratum" to the current doc > without creating a new version, to insert an additional pair of examples > (between the current examples 5/6 & 7/8), illustrating that the prefix > "DC" might be mapped to an example.org "namespace URI" and maybe also > that a prefix "XYZ" might be mapped to a DCMI-owned namespace URI. That sounds very good, if the DC publishing process allows you to do so. > ... > Just to clarify, this question is referring to the profile URIs used in > the profile document itself i.e. in the document > > http://dublincore.org/documents/2008/08/04/dc-html/ > > not in X/HTML documents citing/making use of that profile. Indeed. > Yes, I think you're right. The profile document itself should be citing > both profiles, the W3C GRDDL one and the DCMI one. > > What I think happened is the following: > > The previous version of the profile doc [1] had referred to the Embedded > RDF profile [2] and not the W3C GRDDL profile. The Embedded RDF profile > also defined the rel="schema.XYZ" convention, so the use of that profile > "covered" both the rel="grddl.profileTransformation" _and_ the > rel="dc.[...]" usage. > > Then in preparing the final version we switched to using the W3C GRDDL > profile, rather than the Embedded RDF profile. But that W3C GRDDL > profile only covers the > rel="profileTransformation"/rel="transformation", and not the other > stuff. So essentially the use of @rel="schema.DC" and @name="DC.[...]" > in the profile document is undefined. > > I think I had some doubts about whether there might be a problem with > "recursion" - the profile doc itself referring to the profile it > describes - but I'll run a test on a copy to see what happens. If it > works OK, we'll add the additional profile URI. i.e. In the document > > http://dublincore.org/documents/2008/08/04/dc-html/ > > change > > <head profile="http://www.w3.org/2003/g/data-view"> > > to > > <head profile="http://www.w3.org/2003/g/data-view > http://dublincore.org/documents/2008/08/04/dc-html/"> > > (assuming, as above, that we do this as an "erratum" to the current > document). Yes, that's what I was looking for. Be prepared to get warnings from TIDY-based HTML validators, thouh... (<http://sourceforge.net/tracker/index.php?func=detail&aid=1264455&group_id=27659&atid=390963>). > And again, just to emphasises, there are no new requirements for X/HTML > instances citing the DCMI profile; this would be a change only to the > profile document itself. > > Strictly speaking, I think we should probably also be citing a profile > for the use of rel="meta", as that link type isn't defined as part of > the X/HTML spec. Such a profile is defined here > > http://purl.org/net/uriprofile/ > > So we could add that too? i.e. use > > <head profile="http://www.w3.org/2003/g/data-view > http://dublincore.org/documents/2008/08/04/dc-html/ > http://purl.org/net/uriprofile/"> > > Thanks again for pointing out these two issues, Julian. > ... And thanks for the feedback. In the meantime Leif Halvard Silli pointed me to a related problem: at least some of the documents on dublincore.org use DC-HTML notation, without specifying a @profile at all -- see, for instance <http://www.dublincore.org/documents/usageguide/>. BR, Julian