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

On 2007-09 -29, at 12:56, Dan Brickley wrote:

> Tim Berners-Lee wrote:
>> On 2007-09 -27, at 09:47, Misha Wolf wrote:
>>>
>>> How about 303 redirects?
>> The way we are using them for the Semantic Web, when <A> redirects  
>> 303 to <B>,
>> then <B> is a document describing <A>.    They are not URIs for te  
>> same thing,
>> so you don't have a choice.   The URI to remember for the thing is A.
>
> Would it be wrong for B to be a(nother) URI for the resource? eg.  
> an LSID, URN, java:, phone: or other non-http URI?

The 303 "See other"  redirect I take as meaning "other".     Not  
identifying the same thing.
If you want to tell someone about a new URI for the same thing, when  
it would be
appropriate to use a 301 Moved or 302  Found redirect, depending on  
whether you wanted someone to use the later or the original  URI in  
future reference.  The 301 redirect is a warning message indicating  
an unsatisfactory situation. It indicates the caller used the wrong  
URI.  It should not be designed into a protocol as normal operation.

Now in principle,  it could be all kinds of things,  but to make the  
architecture work, it has to be something useful.    If you redirect  
to a phone number, the only way a machine can figure out why that is  
useful it to call the number, and ask, "Hey, I was just doing a GET  
on http://bio.org/protein/1234 and I got your number".
  So I can't see a 303 as being useful except when <B> is a document  
about <A>.

Once you have done a 303 redirect, or you have client-side stripped a  
hash from A to get a documet about A, then of course you can in that  
document put all kinds of RDF information about A, and about the  
document (like translation, license,  etc etc).  So you don't need  
any more HTTP protocol metadata.

Tim

Received on Monday, 1 October 2007 14:45:35 UTC