[vlc-devel] [PATCH 2/4] httpd: add vlc_chromecast_http_HostNew

[Pierre Ynard]

Hello,

> > > I don't see why this is needed and I don't think Chromecast should
> > > be treated as a special snowflake either.
> > > 
> > > If the user set VLC to bind to a specific interface, everything
> > > should follow the user setting.
> > 
> > I did this patch mainly to let chromecast override the timeout
> > option.
> 
> The timeout option can intrinsically only exist per protocol (or
> globally), not per URL, since that is the only characteristic of the
> connection that is generally known at the time of the evaluating the
> timeout.
> 
> The API is historically kinda broken in that it lacks a layer of
> indirection from physical to virtual host, but the timeout would have
> to be per physical host anyway. I don't see any parameter that's not
> already exposed that would make sense to expose.

I agree that Chromecast shouldn't be treated as a special snowflake. The
API was designed this way for reasons. As Rémi said, there are slight
limitations with this code when reusing hosts; a new one now is how the
new timeout parameter is theoretically ignored when getting an already
existing host. If the limitations are an issue, I think the relevant
thing to do would be to challenge and argue about that API design.

On the other hand, RTSP uses a configuration variable for timeouts, and
at this point Chromecast already overrides some HTTPd host configuration
variables, so I would suggest trying this way to control Chromecast
timeouts too.

-- 
Pierre Ynard
"Une âme dans un corps, c'est comme un dessin sur une feuille de papier."