Trevor Jay wrote: >> On your last point, I understand your reluctance to add *yet another* >> dependency to the camera_stack and probe is very likely to generic to >> be considered a part of a camera stack proper. Still, it is very >> flexible and can be very useful for building imaging toolchains with >> ROS. This should be even more true once the nodelet interface has >> solidified. Blaise Gassend wrote: > For me, the question would be how widespread gstreamer is. If it is part > of all supported distributions, then having a system dependency on it > isn't a problem for me. Even if gstreamer is not available on a system, > that will only cause problems when compiling the package that needs > gstreamer, so only people that need that particular package will have a > problem. My primary concern is that gstreamer would introduce another image stream framework within the ROS image pipeline. That seems architecturally undesirable. In my mind, what we want is more analogous to the dc1394src gstreamer plugin Trevor mentioned. I could be wrong about that and can probably be convinced otherwise. But, that is my current thinking. We already have two good ROS drivers based on libdc1394. I am more inclined to pursue those options. Comments and feedback are (always) welcome... -- joq