That makes most sense to me as well. 

Besides all the non-optimal things I discuss in the internals paper, the 
numbers - of my very limited measurements - show that mod_h2 is slightly less 
performant than plain httpd *if you only have a single request/connection at a 
time*. If you have 2 requests ongoing you break even or are slightly better. 
And then it grows up to 50% more requests/sec than with HTTP/1.1 (all measured 
over TLS). Of course this varies with resource size and such...

Since browsers will typically want to get quite some resources from the server 
for a page, this should already in 2.4 bring benefit for everyone.

cheers, Stefan

> Am 28.05.2015 um 16:46 schrieb Jim Jagielski <>:
> My thoughts are that we use mod_h2 as a guide to how to
> "better" implement things in trunk, but also allow for
> mod_h2 to also work w/ 2.4 as well... So there will be
> a 2.4 version of mod_h2 as well as a more significant
> "merging" of mod_h2/trunk/2.6/3.0.
>> On May 28, 2015, at 10:36 AM, Nick Kew <> wrote:
>> On Wed, 2015-05-27 at 22:42 +0200, Stefan Eissing wrote:
>>> Not wanting to boast, but maybe mod_h2 for httpd 2.4 can play a role in 
>>> motivating people to migrate away from 2.2. 
>> I've just looked at your "internals" page (which seems to me
>> an excellent piece of work), and it tends to support the gut
>> feeling that mod_h2 might be better targeting 2.6/3.0 than 2.4.
>> We should be joining forces to address the issues you've
>> encountered, from minor tweaks to core to more fundamental issues
>> like bucket alloc across threads (or a suitable alternative).
>> Time for me to download and take a proper look at your work,
>> and put my money (or at least time&effort) where my mouth is!
>> -- 
>> Nick Kew

<green/>bytes GmbH
Hafenweg 16, 48155 Münster, Germany
Phone: +49 251 2807760. Amtsgericht Münster: HRB5782

Reply via email to