Re: PUSH_PROMISE and load balancers

Just to follow up on this, in our load balancer implementation at the 
moment we are using an X-Disable-Push header to disable push for a 
specific request.

I would be very interested to hear if any other reverse proxy 
implementors have run into this issue, and if so maybe try and get an 
agreement on a de-facto standard header to disable push for a given request.


Stuart


Daniel Sommermann wrote:
> On 09/26/2014 12:32 AM, Martin Thomson wrote:
>> On 26 September 2014 08:21, Stuart Douglas
>> <stuart.w.douglas@gmail.com> wrote:
>>> I know there are ways to work around it, but it seems sub-optimal to
>>> need to
>>> maintain two separate connection pools for push enabled vs non-push
>>> enabled
>>> clients, especially when it seems to be easy to fix on a protocol level.
>> It's still strictly better than HTTP/1.1.
>
> I don't know about *strictly* better. Two connection pools is an
> increase in complexity in the LB and could lead to worse performance.
> You might have to pay a penalty of opening a new connection to a server
> even though you have a connection ready, whereas in HTTP/1 you didn't
> have to open another connection. Are we really so pressed for bits in
> the flags that we can't move this there? Maybe SETTINGS is not the right
> place for enabling/disabling server push.
>
> Daniel

Received on Wednesday, 10 December 2014 11:45:07 UTC