On Thu, 26 Mar 2015, Andrea Rendine wrote:
> >>  - The properties
> >> dc.xxxx, ...
> >>   are to be REMOVED
> >
> > Are they used by people? If so, it seems like it's the DC 
> > specification that should be fixed.
>
> It isn't the fact whether they are used or not.

Whether they're being used or not is pretty much the most important 
criteria for whether it should be included in the registry wiki page.


> They are associated with <link rel="schema.dc" 
> href="http://purl.org/dc/elements/1.1/";>, which is similar to a 
> namespace declaration. Those values aren't defined in that namespace, 
> nobody used them at the beginning because they weren't defined.

I'm not sure what this means. There's no "namespace declaration" concept 
for <link rel> or <meta name>. The names are opaque and each individually 
defined.


> > I'm not sure what you're saying here. Isn't the DC spec a more 
> > appropriate place for such warnings, though?
>
> The fact is, the first 15 DC.prefixed properties are older and listed in 
> a sort of legacy compatibility. It's probable that some authors have 
> used them in the past, when they were defined (without a proper range, 
> es Agent, Date, Resource, Literal, etc.), so they've been listed in the 
> table as well. But now a more modern and up-to-date namespace exists, 
> defined by the <link rel="schema.dcterms" 
> href="http://purl.org/dc/terms/";> namespace declaration. "Newer" authors 
> who start using Dublin Core metadata should use this declaration, at 
> least if they're going to use properties in their defined range. See 
> http://wiki.dublincore.org/index.php/FAQ/DC_and_DCTERMS_Namespaces

I don't see why this is relevant for the meta wiki. It should just list 
all the terms that people use or could use, linked to the relevant spec, 
with a brief description. That description can say "(deprecated, see 
spec)" or "(commonly misused, see spec)" or some such, where that text is 
a link to a discussion of the state of the art in the spec, of course.


> >> dcterms.collection
> > Is it used by anyone?
> It has no meaning for documents. It defines a datatype. If it is used, it
> is used incorrectly.

If it's used incorrectly 100% of the time that it is used, then the 
problem is probably with the spec, not with the use.


> >> The properties
> >> dcterms.xxxx
> >> are to be REMOVED
> > Again, are they used by anyone?
> Again, the spec says they mustn't, as they're meant to define a <link@rel>
> element rather than a <meta@name>. They'd make no sense when used with a
> string, they refer to a URI.

Ok, but that doesn't answer the question. If they're used, we shouldn't 
remove them, we should fix the underlying spec to say how to handle the 
existing content that uses them.


On Thu, 26 Mar 2015, Andrea Rendine wrote:
>
> According to the spec, the properties marked in my mail can be used as 
> DC-prefixed, though in the wild the DCTERMS prefix is largely preferred. 
> If these properties were to be used, it should be specified that the 
> correct namespace declaration is *not* <link rel="schema.dc" href=" 
> http://purl.org/dc/elements/1.1/";>, but something like <link 
> rel="schema.dc" href="http://purl.org/dc/terms/";>. DC prefix is used 
> with both namespaces for legacy compatibility, but the prevalence of 
> DCTERMS over DC is about 100 to 1.

There's no concept of namespaces here. If the DC spec says there is, it 
should be fixed.


On Thu, 2 Apr 2015, Andrea Rendine wrote:
>
> Mr. Hickson, please let me know if further details are needed, or otherwise
> if change proposal will be accepted, at least partially.
> The more those properties remain there, the more authors can read them and
> use them incorrectly IMHO.

My recommendation is to make whatever edits you think are necessary within 
the following constraints:

 - any values that are used in the wild should be included
 - values that are used in the wild in a consistent manner should be
   specified in the relevant specification (this might require changes to 
   that specification)
 - values that are widly used inconsistently should be obsoleted by the 
   relevant specification and marked accordingly on the relevant wiki
 - the concept of namespaces for these values should be excised from 
   any specification that attempts to introduce it
 - the description in the wiki should be brief, should deep-link to the 
   relevant specification, and should include brief parentheticals for any 
   usage notes, linked to the relevant specification for more discussion

-- 
Ian Hickson               U+1047E                )\._.,--....,'``.    fL
http://ln.hixie.ch/       U+263A                /,   _.. \   _\  ;`._ ,.
Things that are impossible just take longer.   `._.-(,_..'--(,_..'`-.;.'

Reply via email to