W3C home > Mailing lists > Public > www-talk@w3.org > March to April 1997

Re: We need a META schema registry!

From: Andrew Daviel <andrew@andrew.triumf.ca>
Date: Sun, 30 Mar 1997 01:18:51 -0800 (PST)
To: Dan Connolly <connolly@w3.org>
Cc: www-talk@w3.org, meta2@mrrl.lut.ac.uk, swick@w3.org
Message-Id: <Pine.LNX.3.91.970330002957.22279A-100000@andrew.triumf.ca>
On Thu, 27 Mar 1997, Dan Connolly wrote:

> > Given that many organisations will want to use private metadata, and that
> > particular disciplines have their own metadata, a single global repository of
> > all metadata types is clearly unworkable.
> 
> Quite the contrary! Each organization that wants to
> define a metadata field (or schema of fields) need only
> assign it a URL, and (optionally) publish the specification

What I meant was, a single page which is definitive for all possible 
values of META attribute is unworkable. A page which lists URLs of
metadata specifications is more-or-less what I had in mind.

> 	http://www.w3.org/pub/WWW/Search/catalogs#schemas
 .. add to my lists ...

> Your example could be written:
> 
> <META NAME="http://purl.org/metadata/dublin_core_elements#Author" 
> 	CONTENT="Oscar Wilde">

Hadn't seen that variety, but it's interesting.
I'd prefer to see LINK or anchor used with HREF for a URL, simply because
people (and agents) expect that HREF goes with URLs and NAME goes with
text. Something like
<LINK REL=METADATA CONTENT="Oscar Wilde"
  HREF="http://purl.org/metadata/dublin_core_elements#Author">

The more I look at metadata in e.g. GIS systems, the longer I think it's
going to take to unify it (probably never). What I'd like to see is some
kind of middle ground between the hopelessly vague and the academically
precise, that could be generated using an authoring tool by "Joe Sixpack"
and would let me find documents (say) by a particular author without
hitting all the flames, references, reviews and other spurious mentions of 
the guys name. THe Dublin Core has some promise for hitting this spot. 

I did a survey of a few thousand Web pages with a robot at
http://vancouver-webpages.com/META/bycount.shtml
Much of the data is from traversing Yahoo! to a depth of 4.

The most popular NAME META attributes are
keywords, description, generator, robots and author.
The most popular HTTP-EQUIV attributes are
pics-label, content-type, refresh and keywords.
REL and REV attributes on LINK are much less popular.

I guess that most people have a fairly clear idea of what "Author" means, 
and that a lot of people have read the search engine help pages.
Mis-spelling tags has a certain popularity, but the formal schemes are
notable for their absence. 

META attributes are going to be dominated by whatever the authoring and
conversion tools generate. If the authoring tool designers think there's
no standard, we get a hodge-podge of different tags.

Andrew Daviel
Received on Sunday, 30 March 1997 04:46:20 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Wednesday, 27 October 2010 18:14:22 GMT