Comment by adrian_b
7 hours ago
I have already said that it was an USB camera, using the UVC protocol, and that it had FullHD resolution. Nothing else really matters about the interface.
FreeBSD has a dedicated service for USB cameras, webcamd, and it worked very well for capturing video and audio at maximum resolution, and without interference from any other programs that were running concurrently on the server. As I have said, in Linux not only the required configuration was more complex, but I tried several programs and all had stutter problems at FullHD resolution (while other programs were also running on the computer). That was the status at that time. Now, many kernel versions later, I assume that such problems no longer exist, at least not with old cameras.
I do not see what is not tracking for you in this example. It is not an isolated example, for many years FreeBSD was known to have less problems than Linux in handling video streams and audio streams with low latency and constant throughput. More recently, Linux has also improved, but in the past unreliable performance with certain video/audio devices was not unusual (i.e. where other programs running concurrently caused video/audio drops or delays).
That’s fair. I’m struggling to understand how Linux had a harder time interfacing with a USB byte stream than a bsd would. A model for the camera would be great!
I think that it was the Logitech C920, which is still available.
But like I have said, I do not think that the model mattered much. IIRC the camera had an internal video encoder, because otherwise uncompressed FullHD video would not pass through USB 2.0.
The differences between FreeBSD and Linux at that time were at 2 levels. Regarding the user interface, FreeBSD happened to include in the base system programs dedicated for using such a camera, so their use was very simple. On Linux I had to search and install a suitable package, and those that were available were more general video applications and because of that their configuration to do the specific thing that I needed was more complex.
Besides the simpler interface, there was the stuttering problem on Linux, which was caused by the scheduling policies of the kernel. Perhaps it would have been possible on Linux to find a way to ensure a higher priority for the video and audio handling, to not be preempted by the concurrent programs running on the server, but since on FreeBSD everything worked fine out of the box there was no reason for me to investigate how that could be done on Linux.