Metadata naming conflicts

Andrew Daviel (andrew@andrew.triumf.ca)
Tue, 30 Sep 1997 08:32:17 -0700 (PDT)


Date: Tue, 30 Sep 1997 08:32:17 -0700 (PDT)
From: Andrew Daviel <andrew@andrew.triumf.ca>
To: meta2 <meta2@mrrl.lut.ac.uk>, WWW-HTML List <www-html@w3.org>
cc: Udo Schuermann <udo@BlackboardLLC.com>
Message-ID: <Pine.LNX.3.95.970922115026.22560E-100000@andrew.triumf.ca>
Subject: Metadata naming conflicts


Since authors are starting to use metadata within HTML (and other
formats), there is the possibilty of conflict.

Historically, Usenet and later HTML & the major Web search engines have
endorsed the following names:
  Subject
  Keywords
  Title
  Description
which most people seem to understand.

Now that metadata is becoming more prevalent, there is the 
possibility (or rather, virtual certainty) of usage conflict.
 
Ignoring issues of transport - HTML, XML, MCF, RDF etc. - and assuming
that metadata elements may use the same name when transformed to a
different transport mechanism :-

There are various methods of avoiding conflict:
 - Pick an unusual element name
 - Negotiate names in a public forum
 - Use elements only within a restricted domain

For example, Dublin Core (http://purl.org/metadata/dublin_core_elements)
uses two of these methods. Element names within the HTML transport are
prefixed "DC", e.g. "DC.title", and the names have been publicly
discussed. Eastman Kodak uses e.g. "EKdocOwner" within the kodak.com
domain.

With the possibility of global discovery agents using various metadata
elements, elements used within a restricted, but public, domain may still
conflict with elements used elsewhere, since an author may wish to
register pages with, say, both "Education" and "Art" indexing
agents, which may have defined the element "School" as the name of school
(Brent High), and the school of art  (Impressionist)
respectively.

For example, there is a potential conflict in the use of
"Location" between HTTP (redirect URL), Maple Square
(http://www.maplesquare.com/addsite.asp), and IMS
(http://www.imsproject.org/metadata/MDdictionary.html), although currently
these do not use the same transport.

I guess what I'm saying is that I'd like to see new element names
that may be subject to misinterpretation either so thoroughly
hammered out in public that everyone agrees, or else
prefixed with some tag or using some structure that resolves any 
ambiguity. 


I am maintaining a dictionary of HTML META elements at 
http://vancouver-webpages.com/META/, which is currently a list of observed
usage, rather than a definitive guide (though I do take a shot at
people putting their data content in the name attribute).

Andrew Daviel
TRIUMF & Vancouver Webpages