RE: Resource Gloss (Human Readable)

[Charles Reitzel <creitzel@mediaone.net>]

On Sun, 31 Dec 2000, Simon St. Laurent wrote:
>At 11:35 AM 12/31/00 -0800, Tim Bray wrote:
>>At 05:16 PM 30/12/00 -0500, Simon St.Laurent wrote:
>>>Sounds good.  I've got a slightly lighter suggestion.  The document at the
>>>namespace URI could be any well-formed XML, but the only content in it
>>>which matters for purposes of automated processing is simple XLinks.
>>
>>Hm... I'd like to force some aspects of HTML down  their throats
>>to make it clear that it's an expectation that there be some
>>human-readable documentation. -Tim
>
>Fair enough, and I'm utterly sympathetic to that expectation.  I'm just
>pondering how to make this play somewhat nice with the schema-at-the-URI
>approach Rick Jelliffe notes as being in the field already.

1) RDDL looks good, but it directly competes w/ the existing OASIS work
without offering any significant additional benefit.  I'd suggest folks that
need this functionality participate in the existing process, which is well
underway.  The next meeting is Jan 8.  (The discussion about xhtml
extensions has been real interesting, though.  I want to take a closer look
when I get more time).

2) I don't think you can force anything down anyone's throats on this
matter.  The NS spec is clear.

3) There appears to be grudging consensus that you shouldn't URL format
(i.e. protocol identifier prefixes, host and domain, plus a file spec) as a
namespace URI unless it actually works.  It's only truth in advertising.
The RDDL spec doesn't address at all how a general purpose XML processor
should decide whether or not to attempt retrieval.  DISCO uses brute force,
which seems excessive.

Perhaps someone can point me to a good grammar for URL interpretation.  I'd
suggest that a namespace URI be required to match such a grammar before any
such namespace document retrieval is attempted.  I'll go further and say
that the file extension be used to give a hint as to what can be found at
that URL (xsd == XSchema, dtd == DTD, rddl == RDDL, etc.).  If using http, a
Content-type response header would override.

In a perfect world, the W3C would have put their necks out and made some
basic proposal such as this.  They didn't.  XSchema is years waiting now.
Life is short.  You have to break eggs to make french toast.  ... besides,
some people want to use this stuff while still playing well with others.

take it easy,
Charles Reitzel