Re: XHTML2 and metainformation

Tantek Çelik wrote:

> On 5/15/03 3:19 AM, "Robin Lionheart" <w3c-ml@robinlionheart.com> wrote:
> 
> > 
> > Ernest Cline wrote:
> >> Here is a proposal of how to resolve both problems.
> >> 
> >> <ml profile="http://www.dublincore.org">
> >>   <mi name="Publisher">DC Comics</mi>
> >> </ml>
> >> <ml profile="http://www.dccomics.com">
> >>   <mi name="Publisher">Karen Berger</mi>
> >> </ml>
> > 
> > The HTML 4.0 syntax for setting metadata schemas:
> > 
> > <head profile="http://www.dublincore.org">
> >   <meta name="DC.publisher">DC Comics</meta>
> > </head>
> > 
> > does have the limitation of not handling multiple profiles.
> 
> No.  You may have multiple profiles separated by white space:
> 
> <blockquote cite="http://w3.org/TR/html401/struct/global.html#adef-profile">
> profile = URI [CT]
> This attribute specifies the location of one or more meta data profiles,
> separated by white space. For future extensions, user agents should consider
> the value to be a list even though this specification only considers the
> first URI to be significant.
> </blockquote>
> 
> The error in HTML4.01 is that the type is specified as URI, where it should
> be specified as URIs in order to comply with the prose.  This error may have
> been propagated to the latest XHTML2 draft.

No the latest draft does specifically allow for multiple profiles, but 
it provides no way to associate a given piece of metainformation with a 
particular profile. Thus the possibility exists that two profiles could 
establish conflicting methods of association.  While RFC 2731 mandates  
the use of <link rel="schema.*"> and <meta name="*.tag"> to make such 
an association, it has never been part of any W3C (X)HTML standard.

> > The syntax also
> > implies that your "Keywords" and "Description" metas are part of your Dublic
> > Core metadata.
> 
> I don't think neither this, nor a "Dublin Core" metadata scheme is implied.
> 
> 
> > Your proposal is better, but I don't think a cryptic two-letter tag like
> > <ml> is necessary for such an uncommon element, and we already have <meta>
> > so why introduce <mi>?
> 
> I agree with this reasoning.
> 
> 
> > I propose instead:
> > 
> > <schema profile="http://www.dccomics.com">
> >  <meta name="Publisher">Karen Berger</meta>
> > </schema>
> 
> I'm not sure how much sense it makes to specify this level of metadata
> schema detail in HTML.  For instance, is the name "Publisher" in the
> previous example a name global to HTML, or is it defined by
> "http://www.dccomics.com" and then used?

The idea is that "Publisher" is specific to the 
"http://www.dc.comics.com" schema.


> If the latter, and the intent is to attempt to contain all the <meta> that
> uses a particular profile within the declaration for that profile, this
> fails to solve the linktype case, which allows profiles to extend the values
> of the 'rel' and 'rev' attributes, which can/are then used on <a href="">
> tags etc., certainly outside of any metadata schema declaration.

Let me first quote from an earlier post of mine on this subject, since 
its been almost a month since it was made:

# If RFC 2731 is to be adapted so as to be part of XHTML2, then it
# should be referenced normativly in the recommendation.  Currently
# it has only a mention as part of an unreferenced example in the
# Metainformation Module.
#
# Clearly a method of associating metainformation with a
# schema/profile SHOULD be standardized in XHTML2 and incorporated
# as part of the Metainformation Module.  The chosen method is of
# secondary importance.

# There are three methods I see of doing this in XHTML. One would be a 
# minimal rewriting of RFC 2731 so as to adapt it to syntax of XHTML2.  
# 
# This would look something like:
#     <link rel="schema.DC" href="http://purl.org/dc/elements/1.1/" />
#     <meta name="DC.Date">2000-01-01</meta>
# Another would be similar to RFC 2731, but instead of link would use a
# new element defined in the Metainformation Module and look something 
# like this:
#     <schema name="DC" profile="http://purl.org/dc/elements/1.1/" />
#     <meta name="DC.Date">2000-01-01</meta>
# The third would be like my initial proposal:
#     <ml profile="http://purl.org/dc/elements/1.1/">
#       <mi name="Date">2000-01-01</mi>
#     </ml>
# Of these three formats, I really don't like the first as I think that
# associating metainformation with its schema is a task that should not
# be left to <link> but should have its own element. I have a slight 
# preference for the third format, but would have no complaints if 
# something like the second format were incorporated into XHTML2.
# Leaving the association of metainformation with its schema to a
# non-W3C extension is in my opinion totally unacceptable. Even
# sanctioning the first format above and making it normative would be
# preferable.

At the time of that discussion, no mention was made of the desirability 
of using a schema to specify new link types for use with <link> or <a>.
If such an ability is desirable then I agree that the third format
I gave is insufficient.  Of the two remaining formats, I'd prefer 
something like the second where a specific element such as <schema> for 
making that association is used instead of doling that job off to 
<link>, especially if one of the jobs of the this method is to assign 
possible linktypes for <link> to use. Elements that could modify their 
own interpretation make me suspiscious.

Received on Friday, 16 May 2003 00:01:21 UTC