- From: Jamie Lokier <jamie@shareable.org>
- Date: Mon, 1 Sep 2008 10:25:35 +0100
- To: Julian Reschke <julian.reschke@gmx.de>
- Cc: Kris Zyp <kris@sitepen.com>, ietf-http-wg@w3.org
Julian Reschke wrote:
> >interoperability with servers by following the HTTP specification as
> >closely as possible, so servers have a real standard to go off of
> >instead something we made up. It seems like leveraging the range/partial
> >content mechanism with alternate range unit is the approach that HTTP
> >would suggest, and I have no reason to believe it is wrong. Retrieving a
> >paged subset of data is merely a different representation of the same
> >resource.
>
> Yes. But, making up new range units shares has similar problems as
> making up query parameters, doesn't it?
>
> To make this robust, we'd really need a registry.
Yes, a registry or a convention for range-unit namespaces, like
"com.dojo.items", or even "items{uuid=UUID}".
Or simply declare that "items" is application-specific. Only use it
with known resources, and caches should not do anything clever with
it.
Or use a different header: Dojo-Range, with the response containing
"Vary: Dojo-Range". That would make it cachable by generic HTTP
caches, which sounds rather desirable.
-- Jamie
Received on Monday, 1 September 2008 09:26:13 UTC