Hi Harri,

On Monday 25 June 2007, Harri Klapuri wrote:
> Hi Laurent,
>
> >> I know. Actually the problem is the same for stream-based because
> >> the GUID is needed there, too.
> >
> > You're right, but MPEG should still not be a frame-based format :-)
>
> Usually MPEG formats are transmitted as streams but encoders
> and decoders often work on full frames. Forcing the frames
> into a byte-oriented stream seems like a loss of information.

Isn't that information embedded in the MPEG stream anyway ?

> Isn't V4L2 inherently frame-oriented or is there a way to
> pass streams/packets, too? A stream contains frame indicators
> of course, so the driver could be looking for them but that
> means more work for the driver.

V4L2 is buffer oriented. If I'm not mistaken, a buffer is not required to be 
filled with a complete video frame. The driver will return MPEG video data to 
userspace without bothering to check frame boundaries.

> > Has MS defined GUIDs for the different MPEG payloads ?
>
> I am not aware of anything useful in this area.

:-(

> >> It is not out yet (available for sale). Besides the typical MJPEG
> >> and Uncompressed, it supports frame-based payloads.
> >
> > Will it support MPEG as well ?
>
> It is possible. In any case I anticipate the number of
> MPEG4 devices is going to increase pretty soon.

I agree with you there. However, I'd rather wait for the first devices to hit 
the market before adding frame-based MPEG support. As stated before, MPEG 
should be stream-based, so I hope no devices will return MPEG data in a 
frame-based format.

I committed frame-based formats support but removed MPEG support for now. 
Let's discuss the MPEG issue when real-life examples will be available.

Cheers,

Laurent Pinchart
_______________________________________________
Linux-uvc-devel mailing list
[email protected]
https://lists.berlios.de/mailman/listinfo/linux-uvc-devel

Reply via email to