3xx HTTP Response codes

Hi,

We (i.e. the POWDER WG) are getting close to requesting transition to 
first public working draft of our first Rec Track document which looks 
at Grouping of Resources. In it we make specific reference to the  HTTP 
in RDF vocabulary [1] in a couple of contexts. One of these concerns 
redirection. Quick introduction:

The basic model is that a subClass of an OWL Class, known as a Resource 
Set, lists a bunch of things that must be true if a candidate resource 
is to be considered a member of the set. We foresee the following as a 
typical Resource Set definition:

<wdr:ResourceSet>
   <wdr:inclHosts>example.org</wdr:inclHosts>
</wdr:ResourceSet>.

In other words, this is the set of all resources available on 
example.org. Easy enough. But what if, when resolving, say, 
http://www.example.org/index.shtml the HTTP Response is a redirect to 
http://www.example.com/thing.asp - i.e. we've gone from .org to .com in 
this instance. We've resolved that, whilst we could decide definitively 
that the answer is 'yes', there would surely be cases where some people 
wanted it to be 'no' and vice versa. Hence we want to make this up to 
the person defining the set.

So the plan is to say that the default behaviour is no, redirects are 
not elements of the set (unless processing the target of the redirect 
independently shows it to be an element of the set) but that Resource 
Set definitions can include a property to override this so we'd have 
something like

<wdr:ResourceSet>
   <wdr:inclHosts>example.org</wdr:inclHosts>
   <wdr:inclRedirects rdf:resource="http://www.w3.org/2006/http#301" />
</wdr:ResourceSet>

But, as you'll notice, this applies only to 301 redirects which may be 
exactly what is required, so this is useful, however, we'd also like a 
general redirect property like http://www.w3.org/2006/http#3xx - i.e. 
whatever the response code, if it's in the 300s, it's OK.

Now we could, of course, define a class called 'redirect' and say it's 
the union of 301, 302, 303 and 307 (which would be more exact than being 
sloppily general with 3xx) but I wonder whether other users of HTTP in 
RDF might want generic classes like 'success', 'redirect', request 
failed', or 'server problem (i.e. 2xx, 3xx, 4xx and 5xx in very general 
terms)?

Phil.


[1] http://www.w3.org/TR/HTTP-in-RDF/

Received on Monday, 25 June 2007 15:14:27 UTC