> Am 28.04.2016 um 10:08 schrieb Rainer Jung <rainer.j...@kippdata.de>: > > Am 28.04.2016 um 04:30 schrieb William A Rowe Jr: >> On Wed, Apr 27, 2016 at 6:16 PM, Yann Ylavic <ylavic....@gmail.com >> <mailto:ylavic....@gmail.com>> wrote: >> >> I was offline today so couldn't comment on the different messages on >> the subject, so I'll try to summarize (here) my understanding, so >> far... >> >> On Wed, Apr 27, 2016 at 8:41 PM, <wr...@apache.org >> <mailto:wr...@apache.org>> wrote: >> > Author: wrowe >> > Date: Wed Apr 27 18:41:49 2016 >> > New Revision: 1741310 >> > >> > URL:http://svn.apache.org/viewvc?rev=1741310&view=rev >> > Log: >> > >> > Ensure http2 follows http in the meaning of >> > status WRITE (meaning 'in the request processing >> > phase' even if still consuming the request body, >> > not literally in a 'now writing' state). >> >> This is indeed consistent with how we report http1 state currently, >> but maybe it could be more intuitive to report READ until the body is >> consumed in http1 rather than changing http2? >> Unless we want to minimize scoreboard updates, for performance >> reasons... >> >> >> Well, we always want to be considerate of performance. >> >> That said, we absolutely must not change the semantic meanings of >> the server-status results on the maintenance branch. Change those >> meanings on trunk for a future 2.6/3.0? Sure... but 2.4.x needs to >> behave as 2.4.x has behaved from the start. > > I agree we shouldn't change the semantics of R/W during 2.4. People might > monitor the numbers of R and W and will be quite surprised by seeing a > noticeable shift from W to R. > > I think it would be a good change for 2.6/3.0 to make R including reading the > request body and only after having read it switch to W. We can mention it in > CHANGES / migration guide and personally I think those semantics would be > more intuitive.
That is how I understood it to be when I made the status updates in mod_http2. But it should of course be in line how we report the http/1.x scores.