norm's RDDL action item

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

I took an action[1] to summarize the various RDDL designs submitted in
response to Tim Bray's RDDL challenge. Looking through the archives, I
identified six proposals by Tim Bray[2], Chris Wilper[3], Tim
Berners-Lee[4], Jonathan Borden[5], Micah Dubinko[6], and Sandro
Hawke[7], respectively.

Herewith is my attempt to summarize the proposals.

Broadly speaking, they fall into three categories:

1. Some proposals are RDF (4, 5)
2. Some proposals are "valid" HTML (possibly with minor changes to HTML) (6, 7)
3. Some proposals use mixed namespaces but not specifically RDF (2, 3)

The RDF proposals have the benefit of being immediately useable in RDF
metadata frameworks. There are many different ways to express the
relationships in RDF, so it isn't clear that saying "it should be in
RDF" does very much to narrow the scope of syntax discussions. (If
anything, it muddies the water because it exposes all of the RDF
syntax options and that's likely to be distracting.)

It's also pretty clear that any of the non-RDF syntaxes could easily
be harvested into RDF without ambiguity. So if there's a case to be
made for a simpler, easier to author syntax, and if that syntax isn't
RDF, that doesn't mean the data can't easily be transformed for use in
RDF engines.

The "valid" HTML proposals come in two forms. One[6] suggests that the
<meta> tag could (and perhaps will be) changed in XHTML 2.0 to allow
content. This would allow nested <meta> elements to express the
relationships.

One of the nice things about RDDL is that it allows you to mix prose
descriptions and machine-understandable links. I'm concerned that
moving all the machine-understandable stuff up to the head of the
document may force the human and machine bits to be separated, leading
to harder to understand, harder to maintain RDDL documents.

The other proposal[7], extends the "rel" and "type" attributes on HTML
anchors so that they can be directly annotated to indicate that
they're identifying RDDL natures and purposes. That looks like a nice
low-cost solution if HTML validity is important.

The mixed namespace proposals just invent a vocabulary for RDDL links
and embed it in the RDDL document. One[3] puts them all in the head,
about which I've already expressed reservations, the other[2] simply
sticks them in the content.

These proposals have the nice feature that they stand out to a human
reader. "The elements that begin "rddl:" are the RDDL elements."
There's no structural or attribute value subtlety.

In closing, I observe somewhat unhelpfully that what we're talking
about here is a vocabulary of about five or six things. It's so small
that I'm confident we could explain any of the syntaxes to most web
content managers in the space of a lunch break and they'd grok it well
enough to start using it that afternoon.

I think we need to accept that they're all basically equivalent and
pick one.

"So pick one, norm", I hear you say.

Ok. I pick Jonathan Bordan's straight-forward RDF implementation[5] of
Tim Bray's original proposal[2]. But if pressed to vote, I'm voting
"concur". :-)

                                        Be seeing you,
                                          norm

[1] http://www.w3.org/2002/11/18-tag-summary
[2] http://lists.w3.org/Archives/Public/www-tag/2002Dec/0048.html
[3] http://lists.w3.org/Archives/Public/www-tag/2002Dec/0056.html
[4] http://www.w3.org/2002/11/rddl/
[5] http://lists.w3.org/Archives/Public/www-tag/2002Dec/0099
[6] http://lists.w3.org/Archives/Public/www-tag/2002Dec/0180.html
[7] http://lists.w3.org/Archives/Public/www-tag/2002Dec/0236.html

- -- 
Norman.Walsh@Sun.COM    | Talk as if you were making your will: the
XML Standards Architect | fewer the words the less the
Web Tech. and Standards | litigation.--Gracián
Sun Microsystems, Inc.  | 
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.0.6 (GNU/Linux)
Comment: Processed by Mailcrypt 3.5.7 <http://mailcrypt.sourceforge.net/>

iD8DBQE+FhAKOyltUcwYWjsRAkACAJ9TzTow3RP8CHaxQE+AwkIvxoGPNQCfdwEG
orZ0+8GcIB0js11MLrCfRdk=
=EWRp
-----END PGP SIGNATURE-----

Received on Friday, 3 January 2003 17:36:55 UTC