Re: [GnomeMeeting-devel-list] Re:[GnomeMeeting-devel-list] Re: [GnomeMeeting-devel-list]Re:[GnomeMeeting-devel-list] Re: [GnomeMeeting-devel-list]possible PAT

[Damien Sandras <dsandras seconix com>]

Le lun 15/09/2003 à 11:15, PUYDT Julien a écrit :
> Sloppy argument. Gnomemeeting isn't "all apps". It has specific ways to
> use the devices. The point of testing the V4L1 plugin is to check
> whether the single-buffer case can handle all gm needs. If it doesn't,
> then yes, it is valid to implement it as a workaround for incomplete
> drivers.

You will never be sure it always work just by asking people to test CVS.
People testing CVS is a very limited subgroup.

> 
> >  and also performance-wise.
> 
> Does gm really _needs_ 3000fps ?! There's a difference between xawtv and
> gm: one just puts the pics on display, and wants to do it as fast as
> possible. The other wants to send the pics through a network, and hence
> won't want that many pics...
> 

No need to slow things at the hardware level if we know that it will
still be slowed down later at the network and encoding/decoding levels.


> Well, in fact, rename "V4L1" to "QuickCam drivers are crap", and we have
> a plugin that will only handle those cams ;-)
> (yes, that is another solution: avoid polluting the main V4L plugin by
> pushing buggy drivers to more specific plugins!)

I repeat that the right fix is to check the minimum of buffers
available. By reducing things to one buffer for all plugins:
- you take the risk to potentially break several drivers because of
qce-ga
- you loose performances

It seems more logical to me to take the minimum risks and only use one
buffer when we are forced to. Seeing the size of your patch makes me
confortable to think that it is pretty trivial to do.
-- 
 _	Damien Sandras
(o-	GnomeMeeting: http://www.gnomemeeting.org/
//\	FOSDEM 2004:  http://www.fosdem.org
v_/_	H.323 phone:  callto:ils.seconix.com/dsandras seconix com