RE: Which URI should be persistent when redirects are used?

On Thu, 2007-09-27 at 15:21 +0100, Misha Wolf wrote:
> OK.  The story is ...

Ah... good to have more context... I think I see what's
going on up to...

> The IPTC is likely to approve the NewsML-G2 specification at our 
> meeting in Prague in mid-October.
> 
> We use URIs to denote concepts associated with the News Object (as
> values of Subject, Genre, Creator, Title, Publisher, etc).  We 
> want these URIs to be invariant and to participate fully in the 
> Semantic Web.  If a receiving system uses HTTP to access 
> information about a concept (such as Jazz, Marcel Marceau, Mona 
> Lisa, Oil, Reuters), the system hosting the taxonomy in question 
> may have multiple representations of the taxonomy, eg:
> -  RDF/XML
> -  IPTC Knowledge Item
> -  Web pages in various languages
> 
> So let's say that Mona Lisa has the following URIs:
> -  the one used in News Objects
> -  the one which will deliver the RDF/XML
> -  the one which will deliver the Knowledge Item
> -  the one which will deliver the Simplified Chinese Web page
> -  the one which will deliver the Traditional Chinese Web page
> -  the one which will deliver the Japanese Web page
> -  the one which will deliver the International English Web page
> -  the one which will deliver the US English Web page
> -  etc
> 
> One could then write a set of assertions along the lines of:
> -  URI-Y refers to the Simplified Chinese Web page describing URI-X
> 
> And it would be a Good Thing if URI-Y were invariant.
> 
> But (it seems to me that) it would be an even Better Thing if URI-X
> were invariant.

I'm not sure what you mean by "invariant"?

I suggest you choose a URI like
  http://well-established-museum.example/great-works/mona-lisa#it

for the mona lisa, then 
  http://well-established-museum.example/mona-lisa
is naturally a document about the mona lisa, and
you can use content negotation in the usual way, by
storing files called
	mona-lisa.html.en
	mona-lisa.html.ja
	mona-lisa.rdf
	mona-lisa.newsml

on a suitably configured apache server.

If you're nervous about the pun between #it referring
to a section of an HTML document or a painting
(i.e. the open TAG issues fragmentInXML-28 and
RDFinXHTML-35) you can give a 303 redirect to the .html version
in response to GET requests to /great-works/mona-lisa
from clients that seem to prefer HTML.

I'm hopeful that we'll change the MIME spec for text/html
to address this pun before too long, but opinions
on the best way to go vary, even within the TAG.

p.s. this arrangement of doing a redirect to the HTML
version is something Henry Thompson suggested
in discussion of the XML Schema namespace document.

Hmm... I think there was a whiteboard discussion and
somebody took a photo, but I don't see it in the in-progress
minutes...
http://www.w3.org/2001/tag/2007/09/18-tagmem-minutes.html#item04

-- 
Dan Connolly, W3C http://www.w3.org/People/Connolly/

Received on Thursday, 27 September 2007 15:14:22 UTC