> Am 22.07.2015 um 22:38 schrieb Yann Ylavic <[email protected]>: > > On Wed, Jul 22, 2015 at 9:37 PM, Stefan Eissing > <[email protected]> wrote: >> >>> Am 22.07.2015 um 19:00 schrieb Yann Ylavic <[email protected]>: >>> >>> On Wed, Jul 22, 2015 at 6:42 PM, Stefan Eissing >>> <[email protected]> wrote: >>>> So, a RUN_ALL that collects willing protocols pre_select, a ALL for >>>> post_select >>> >>> post_select is to announce the selected one, right? >>> That looks good to me. >> >> Ok, I'll start something tomorrow. >> >>>> and a FIRST for switch? >>> >>> Isn't the core doing the switch itself, after post_select? >> >> Two reasons for this: >> 1. the post_select should run for all before anything happens >> 2. switch on Upgrade headers must be handled immediately by a hook as the >> server has just send the 101 and the switch must write the response in the >> new protocol. > > Ok, so the first (hook ordering) registered module and "aware" of the > new protocol would handle it? > Maybe could we rather do that in the post_select hook (RUN_FIRST), and > prevent any other/later module from thinking it could be the one?
Having slept about it, I see your point. Then it is the double hook (all/first).
