On Wed, Nov 15, 2006 at 09:22:05PM +0200, Oded Shimon wrote:
MPEG is often used in closed systems, or in systems will very well defined constraints. I see no reason why (a future, complete) NUT wouldn't be suitable as a base format in such systems.
well, is the dts ordering the only thing preventing use of nut in such a system? would addition of a delay field to each stream header which would then be added to all timestamps in that stream solve the issue? would a maximum 250ms on such a field be enough? and some additional rule like delay MUST be 0 unless the file is encoded for a specific specification requireing a larger delay?
BTW, I'm against this - if nothing else, we spent a lot of time calculating and figuring out the syncpoint stuff to make sure exact seeking always works. If we did this, we'd have to revisit the issue and make sure nothing breaks. BTW, a fake large decode_delay for the audio stream causes an audio preload. but it is ofcourse illegal.
Any nonzero offset between streams is absolutely totally out of the question for reasosn I described in other posts. There's no need to even consider this topic. Can we drop this useless thread? Rich