Re: Proposed issue: site metadata hook

I've been watching this discussion pretty closely since exposing large amounts 
of metadata is a problem I'm currently trying to solve for a research project. I 
thought I'd just throw some thoughts out there.

One, I really don't see the need for a standardized metadata retrieval mechanism 
because I can think of a thousand and one different ways that I'd like to 
retrieve and process 'metadata'....

Still, just a few comments on Roger's and David's proposal: I think it's pretty 
kludgy. I think HTTP headers were meant to carry information about the message 
(eg Content-Type, Content-Length) or request contextualization information that 
might change how the request is processed (ie content negotiation). I don't 
think using headers in this manner (ie "here's an extra piece of information you 
_might_ be interested in) is good design, nor is it really scalable (surely 90+% 
   of users just won't care about metadata) and it shouldn't be encouraged.

 From a more practical point of view, I'd hate to see people start deploying 
metadata aggregators that go around GET'ing resources, throwing away the 
resulting data all so they can see one precious header.

Also, I'm not sure what's so special about RDF. It seems like the entire problem 
of 'metadata retrieval' is just a denegerate case of the real thorny problem of 
service discovery. I come to a website and I say "I'm looking for documents of 
'this type', can you help me out?" And, IMHO, service discovery is one of those 
things  that it's just impossible to please all the people al the time...

Also, before we go about introducing such fundamental changes that'll require 
big code alterations I'd like to suggest people start looking at more efficient 
and 'low-tech' solutions particularly XML processing instructions and the HTML 
'link' tag. These two solutions are already implemented and people are using 
them right now: a lot of service discovery, from stylesheets to RSS to i18n, is 
happening right now with these solutions and I don't think they should be 
abandoned for "boil the ocean" schemes like new verbs or new headers.

It seems to me that you only really need to discover one 'metadata file' for 
each server. This file could either contain all the RDF that the server 
administrator wants to share or it could contain a bunch of links to all the 
metadata documents the server admin thinks clients might be interested in.

- Buko O.

Received on Tuesday, 18 February 2003 23:41:40 UTC