From an NWC Reader
Here's an interesting letter we received on our Ask the Experts forum today. Normally, we receive very quick and pointed questions (why won't my router work; how do I set up a CIFS share, and the like). But this question was both intriguing and puzzling. If you're into streaming media and SMIL, check it out and lend a hand, if you can.
Network Computing Experts:
We appear to have discovered a bug in Quicktime''s support of SMIL and the Sorenson 3 codec.
We find that when we use the "clip-begin"/"clip-end" SMIL parameters to control quicktime movies encoded with Sorenson 3 we experience extreme visual playback problems.
Try this script on any movie compressed in Sorenson 3.x:
<smil>
<head>
<layout>
<root-layout width="320"
height="240"
background-color="black" />
<region id="full"
title="mediaSegment"
fit="hidden"
left="0"
top="0"
width="320"
height="240"
z-index="0" />
</layout>
</head>
<body>
<seq>
<video src="rtsp://yourserver.yourdomain.com/someSorenson3Movie.mov"
region="viewer" clip-begin="smpte-30-drop=00:00:15:09"
clip-end="smpte-30-drop=00:00:30:25" />
</seq>
</body>
</smil>
Note that the code itself works fine, video playback begins and ends at the specified points, it is only the quality of playback within the first second or two that is unacceptable.
We[base ']re using Quicktime format video, encoded with Sorenson 3 (we have also tested Sorenson 3.1), our target player is Quicktime 5.0.2, the behavior has been confirmed on Windows (IE, Nav and standalone Quicktime player) and Macintosh (standalone Quicktime player).
The behavior can be described as an omission of all key frame information. The only pixel blocks which are displayed are those that are changing; everything else is gray. Once the first key frame is reached, playback proceeds normally. Simple observation of this behavior leads us to believe that only delta frame information is being displayed between the clip-begin and the first key frame.
The hypothesis that key frames are involved is further supported by experiments in which we increased the frequency of forced key frames and observed shorter periods of distorted video.
Interoperability of QuickTime and Sorenson 3 is suspect because this behavior is not observed in movies encoded in an earlier version of the Sorenson codec. Unfortunately it appears that this older codec will not be able to render video within acceptable quality and bandwidth parameters for us to use it as a workaround.
Can anyone confirm this behavior and/or put forward any ideas for a solution?
Thanks,
Dan Steinhauser, Software Engineer
Ignite!, Inc.
http://www.ignitelearning.com
Posted by Brad Shimmin at 10:37:09 AM
|
