Re: Push and Caching

I think pushed responses must not use "no-cache" in Cache-Control. Then 
there is no revalidation problem.


Am 26.08.2014 um 08:48 schrieb Mark Nottingham:
> "fresh on the origin server" isn't relevant; what's relevant is whether they're fresh in the cache, and that can be determined by examining the response.
> The issue at hand is whether the pushed response needs to be revalidated, as per the definition of no-cache.
> On 26 Aug 2014, at 3:49 pm, Matthew Kerwin <> wrote:
>> On 26 August 2014 15:18, William Chow <> wrote:
>>> Can "fresh" work? I agree that it perhaps implies caching as well,
>>> but at least it avoids the notion that the server actually performed
>>> any validation (which it could not, without the client providing
>>> validators for the pushed responses).
>> "Pushed responses are considered fresh on the origin server (...) at
>> the time that the response is generated." Makes sense to me, although it
>> starts to sound a bit no-brainish.
>> ​And regarding your other question:​
>>> Also, which response is the point of reference for
>>> validity/freshness? The proposed sentence seems to refer to a pushed
>>> response being "validated" at the time that the pushed response
>>> itself was generated. I assume we'd actually want to treat the pushed
>>> responses to be fresh at the time the response for the
>>> associated/original request was generated.
>> It can only be fresh at the time the pushed response itself is
>> generated, surely. The original response triggered the *need* for the
>> pushed resource, but there's nothing stopping the value of that pushed
>> resource changing between the need being determined and bytes being
>> transmitted.
>> --
>>    Matthew Kerwin
> --
> Mark Nottingham

Received on Tuesday, 26 August 2014 11:31:35 UTC