On Wed, Dec 28, 2016 at 3:56 PM, Eric Covener <cove...@gmail.com> wrote:
> On Wed, Dec 28, 2016 at 9:54 AM, Yann Ylavic <ylavic....@gmail.com> wrote:
>>> I guess we just got unlucky when this overlap was "fixed" before since
>>> the order is not deterministic. I don't think we'll break anyone by
>>> making wstunnell try first.
>>
>> I'm not sure the different schemes ("https" vs "wss") would allow both
>> handlers to run successively.
>>
>> r->filename is built once for all (proxy_trans), so mod_proxy_http
>> would DECLINE "wss://" and same for mod_proxy_wstunnel with
>> "https://";.
>>
>> I proposed Upgrade handling in mod_proxy_http (a while ago, see [1]),
>> which is IMHO a better way to deal with HTTP Upgrading in httpd...
>
> Doh, I guess at least having the order be deterministic is not harmful
> so I will leave in r1776290Yes, makes sense to try WebSocket first. Maybe we should 426 (Upgrade Required) instead of DECLINing at AH02900, though.
