[Pxc] pxc and pxcdd libraries anyone ?

Alessandro Rubini rubini@gnu.org
Sat, 9 Sep 2000 00:50:05 +0200


> I know, some things suck, some just don't work (e.g. exporting a BMP image),
> but to me the overall design seems sound - especially compared to other
> frame grabber SDKs I've seen. 

Well, I've no such experience, I've never written anything under windows.
 
> For example, I like the simple way the API implements
> asynchronous grabbing, without forcing you to use some ring
> buffer structure and associated locks as others do. 
> It's as simple as

Ok, I don't have similar functionalities (yet). 

> What exactly you don't like?

Name space pollution, for instance. I didn't understand all the
methods are fields of a grabber object ("pxc->Grab()" is good, while
"Grab()" is not). And other stuff that presumes exclusive access to
the device or too PC-specific. But, as I reported, I didn't ever use
that stuff.

> Yup, a library is exactly what I have in mind. It sits on top of your
> driver and provides a drop-in replacement for Imagenation's higher-level 
> APIs.

... as long as the low-level functionality is implemented :)

> Again, I am not quite clear about the compatibility bit - compatible 
> with what other library? 

the dos/windows ones.
 
> Does your application use an X server extension (MIT-SHM?), or do
> you just copy the pixels into a pixmap and pass it to the server?

Nothing at all. I don't like graphics programming and have no
experience with it. My live video stuff just uses an image widget and
tells it to read a pgm or ppm file from /dev. Nothing fancy but tool a
few minutes to implement.

I you want real live-video, you should use video-for-linux. But
writing/maintaining such a mixed grab/display driver is a nightmare,
not something I can afford at all. I went my way since I needed
features that were not available in the video-for-linux
implementation.

While my driver is not as full-features as it may be, I now have
another person looking over it, so some glitches will be sorted
out and mid-implemented features will finally see the light.

/alessandro