[Red5] Major Red5 bug

[Jeremy Morton]

On Tue, 14 Apr 2009 02:02:57 -0700, "Andy Shaules" <bowljoman at hotmail.com>
wrote:

You know, it'd be nice if you didn't adopt such a confrontational attitude.

> Hmm,
> ok
> 
> ffmpeg, xuggler, and probably I assume VideoLanClient for source?

No, I'm using a Sercomm IP camera for the source MPEG4 video.

> I also assume you have posted the same meesage on those mailling lists?

I've discussed this stuff in detail on the xuggler-users list.

> I know one thing in streaming technology.
> 
> this almost never works...
> 
> timestampCount += this.tsInterval;
> 
> If ffmpeg fails to decode data, you better add that frame ts interval
too.

All that means is that the odd frame will be lost, so if I buffer 4 seconds
of data @ 10FPS, after 40 lost frames, the client will need to rebuffer. 
Acceptable to me, and I don't think anything to do with the 'buffer
suddenly resets to zero' bug I've been talking about.

> So to finish this off.
> 
> There is no major red5 bug. period... endo of topic.

Why, because you assert it?  I don't think this has been looked in to fully
yet.

> All it took for me was one search in google...
> 
> 'rtp ffmpeg bug'
> 
> Seems you have a problem even befor you get to red5.

I'm trying to find out the *specific* reason that the buffer is resetting
(if I don't do a Netstream.pause/resume) when the buffer is full.  Your
'you have a problem before you get to red5' statement is nebulous, at best.
 The best theory so far seems to be my lack of sending any audio data.  If
you really have to give audio data to Flash to get it to work, that's
pretty depressing... although obviously it's possible to hack it in to
working using the pause/resume thing.  The other point i'd make (for the
millionth time; everyone seems to be ignoring this) is: THIS BUG DOES NOT
HAPPEN IN FLASH 9!!  It seems to have started in Flash 10.  So, apparently,
something has changed that perhaps now requires audio data to be able to
sync properly, which is a shame.  I was kind of hoping there'd be something
else that had changed, like handshaking or something, which Red5 could do
differently, to fix this.  I put it to this list in the hope that people
who knew something more about Red5 would actually try to help analyze
exactly what was happening with Flash, instead of just continually
asserting that there was a problem with my code.

> By the way. it may seem ludicrous to control timing by the number of
audio 
> samples in a frame, but guess what...
> 
> Windows multimedia playback does the same thing.... AND they do it by 
> changing the system clock to allow continuous playback of a stream when 
> samples do not add up to some frame rate.
> http://msdn.microsoft.com/en-us/library/dd389161(VS.85).aspx
> 
> Interface ActiveMovie ClockAdjust

Heh.  Oh well, it's not like MS are a beacon of good practice.  Otherwise I
might be streaming my stuff in WMV.  :-)

Best regards,
Jeremy Morton (Jez)