[AG-TECH] Re: AG video node using berkley mjpeg work

[Andrew Swan]

Matthew Delco writes:
> > If I set the mjpeg capture rate greater than 10 mbps the display machine
> > never saw the mjpeg stream.  I don't understand why this is true.
> 
> The only thought that comes to mind is that vic doesn't have the ability to
> recover from data loss (unless a special compile-time flag is set in
> openmash).  I imagine that beyond 10mbps a network packet/frame gets dropped
> from each video frame.

what is the compile time flag?  i found no ifdefs in
decoder-jpeg.cc...  if packets are being dropped in the way
that matt and i mentioned, it will definitely be reported
in the packet loss (the percentage in parantheses next to
the source name in the thumbnail window).  bob, when you
see the bandwidth go down, does the loss rate go up at the
same time?

> > On the Display machine (Win2K), when I expanded the mjpeg vic window,
> > the reported frame rate dropped from 30 fps to about 15-17 fps.  The
> > data rate reported also dropped from 10 mbps to about 6-7 mbps.
> 
> What would be nice is if vic decoded each network frame as it receives it.
> However, vic waits for a whole frame to be received, and then decodes the
> entire frame.  During this decode process, the OS buffers are filling up
> and data eventually gets dropped.  This causes more jpeg frames to get
> dumped, which (from what I understand) causes the perceived bitrate
> displayed in the GUI to decrease.

i don't understand why you think it would be good to decode on
a packet-by-packet basis. it would definitely complicate the
implementatoin of the decoder to decode frames fragment by
fragment...  to keep the socket buffers from overflowing, the
JpegDecoder object should be able to drop frames if it can't
keep up.  in any case, the default size for the receiver buffer
(80kb) seems a little low -- bob and others, you might try
bumping it up a bit which at the moment can only be done by
editing net/net-ip.cc directly.