RE: "Variant" language in Content-Location (Issue 109)

Henrik Nordstrom wrote:
> > Is there a need to express the client's expectation that 
> > the "Content-Location" URL will only have a single entity
> > associated with it? Or is a client expected to follow
> > "Content-Location"s for a  number of times?
> 
> It's a protocol expectation as well, very visible in the 
> cache model, so yes if people really think this can be 
> misunderstood then yes it should be clarified that the 
> Content-Location URI SHOULD NOT be a negotiated resource.
> 
> (it doesn't need to be a MUST imho)

Because the cache doesn't dereference the resource identified by the
Content-Location, it doesn't really matter if that resource is negotiated or
not.

To optimize the performance of caches, there are really two things you need
to do: (1) Multiple variants of a resource SHOULD NOT have the same
Content-Location (to prevent unnecessary invalidations), and (2) each entity
(representation) of each variant of a resource SHOULD have the same
Content-Location (to cause invalidations to happen as soon as possible).

Regards,
Brian

Received on Thursday, 7 August 2008 19:58:28 UTC