seek seems to use DTS fields for a lot of its internal logic, while it's probably more in line with what the user wants to consider the PTS instead. Also, as it stands, seek adjusts the seconds argument by the first DTS in a stream. I don't think this makes that much sense to do, especially in a container with multiple streams, such as a video with audio, as it would be different for each stream. Instead of adjusting the zero reference time while seeking, why not use the same reference as both the container and the stream?
seekseems to use DTS fields for a lot of its internal logic, while it's probably more in line with what the user wants to consider the PTS instead. Also, as it stands,seekadjusts the seconds argument by the first DTS in a stream. I don't think this makes that much sense to do, especially in a container with multiple streams, such as a video with audio, as it would be different for each stream. Instead of adjusting the zero reference time while seeking, why not use the same reference as both the container and the stream?