httpRange-14 review

I had an action item to review Dave Wood's and Dave Booth's messages on 
httpRange-14.

I've first looked at Dave Wood's use case [1]. I agree with Dave Booth's 
remarks in [2].  In addition, I think the purl.org example at the end of 
the use case could be potentially confusing because it concerns a 
'hashless URI' that gets 303 redirected to a 'hash URI' with a fragment 
identifier.  I do like, however, the idea of taking the widely known 
dc:creator URI a an example, and I do not think the '#' is a real issue 
because it is not in the URI that gets redirected. Still, it could be a 
source of confusion.  Also note that purl.org currently returns a 302 
and not a 303 redirect, as stated in the example.

Secondly, I looked at [3] about minting URIs.  I basically agree with 
Dave Booth's analysis, but I think the tone is too negative about the 
use of hash URIs, that are widely used.  Strictly, the document can be 
read as endorsing hash URIs in case it is clear that the URI refers to 
an RDF or OWL document, but the overall feeling I was left with at the 
end of the document was "DON'T USE HASH URIs! THEY ARE EVIL!".

I would prefer a conclusion saying that for SemWeb applications, hash 
URIs are OK and common practice in cases where the RDF/OWL context is 
clear, as in
   http://www.w3.org/1999/02/22-rdf-syntax-ns#type
but that 303 redirects are needed when the target is of another MIME type.

The document also states:
    However, if the returned HTTP media
    type is unrestricted (e.g., if it might be HTML)
I do not see a clear use case where the person that is minting the URI 
has to deal with the issue of an unrestricted MIME type: if she has the 
power to choose the hash or 303 option, she most likely also has the 
power to control the MIME type. Or am I missing something here?

When looking at the two messages combined I think it would be very 
usefull to have a single, short document that briefly explains:
- the consequences of httpRange-14 for SemWeb applications,
- the pros and cons of hash and 303 http URIs,
- the use of non-http URIs like lsid,
- and points the user to Alistair's cookbook [4].
It would also be nice to have a short overview exlaining how very common 
namespaces such as rdf, owl, dc, foaf etc deal with this issue.  The dc 
practice to use hashless URIs that point to a 30x redirected hash URI 
might be actually be a kind of best practice we might want to endorse as 
a WG.  But this would change the scope of drafting a simple TAG reply to 
editing a WG Note ...  But it may be worth the extra trouble.

Jacco

[1] http://lists.w3.org/Archives/Public/public-swbp-wg/2005Sep/0010
[2] http://lists.w3.org/Archives/Public/public-swbp-wg/2005Dec/0055.html
[3] http://lists.w3.org/Archives/Public/public-swbp-wg/2005Dec/0056.html
[4] http://www.w3.org/2001/sw/BestPractices/VM/http-examples/

Received on Monday, 12 December 2005 20:36:31 UTC