Re: Indexing non-HTML objects

On Sat, 3 May 1997, Gregory J. Woodhouse wrote:

> Perhaps I misread your post, and so I'm just stating the obvious, but the
> existence of a LINK tag in HTML really has little to with the existence of
> a LINK method in HTTP. Of course, it would be desirable that the server be
> aware of HTML LINKs, so it could process link data without having to
> parse the file. But otherwise, links at the HTTP level are quite
> independent of HTML. 


I realize there's no formal tie between HTML and HTTP. However, as the 
elements have the same name, and the same vaguely-worded suggested use, 
it seemed reasonable to use them for the same thing.


There is a Metadata proposal now in WebDAV, involving new commands in 
HTTP/1.1. Also, there's the PICS-ng work (as yet 
unpublished, I think), and Apple's MCF, plus doubtless other work I'm not 
aware of. I was trying to propose something that:
 - would work with existing HTTP/1.0 servers
 - would work with existing browsers
 - was easy to understand and implement
 - didn't involve changes to existing standards

I think that there's always going to be more authoring tools than 
browsers, ranging from vi and notepad, through formal HTML generators, 
format convertors, dynamically generated content, database backends etc.
This makes the widespread use of complex metadata less likely.
HTML works so well in part because it's simple, and browsers are tolerant
of many syntax errors.

Defining overly complex metadata schemes in my view is a surefire route 
to limited penetration, unless you've got a motivated group of authors 
who have to meet certain standards for a particular publication 
(such as an academic journal). 

The MathN broker system at 
http://www.mathematik.uni-osnabrueck.de/preprints/shadow/
seems an excellent indexing scheme, but the metadata is formally tied to
the resource by a Dublin Core element, which may be too heavy for some 
people. I suggested  LINK with the legacy keywords, description, 
author metadata (or just plain text) as a lighter version.


Seems to me people using a JavaScript generated content have the same 
problem as those using frames - think carefully about what keywords
and description you want, then include them as metadata. Kind of messes 
up fulltext searches, though.

Andrew Daviel

Received on Tuesday, 6 May 1997 12:58:16 UTC