rtsp:// files should default to a lower cache than 8 Mbytes!

[mplayer@…]

When loading a realaudio stream via:

mplayer rtsp://rmv8.bbc.net.uk/worldservice/offtheshelf5_v8.ra

The cache size defaults to 8192 kBytes. This is somewhat oversize! The result is
that the playback takes *ages* to start, even on a fast connection, especially
if the server is throttling the stream (as the BBC do).
100 kByte might be a more reasonable size.

[reimar]

Read the manual!
Have a look at the -cache -cache-prefill and -cache-min options!
Well, the last two seem to be undocumented at the moment, they (more or less)
specify how much percent of the cache must be filled before starting playback.

[mplayer@…]

Thanks for your reply. I did read the manual, and I know about the -cache
option. But I think that mplayer could be smarter by default. Also, that manual
is so complex that it confuses me sometimes: a relative newbie would never find
the option!

Richard

[r_togni@…]

There is no easy way to do it.
As you can see looking at MPlayer console messages (use -v to get more), cache
is filled before file is parsed by demuxer, so MPlayer knows nothing about the
file until cache is not filled enough to start playback.
Cache is started automatically for network streams (else you have to use
-cache), and size depends on stream type. Real rtsp demuxer have quite a big
cache because it's needed for video streams, and MPlayer has no way to tell that
stream is audio only before playback starts.

[mplayer@…]

Thanks for your comment. If it can't be done. I guess that's fair enough.
How about the following though:

if ((the cache is not nearly full within 20 seconds) AND (no cache size has been
specified on the command line)); then start playing anyway.

That should catch the problem of things like the BBC radio on demand service
which is seriously throttled, and will only load at a slow rate (i.e. it's not
friendly to downloading lots of it in advance).

But otherwise, I guess that this is a wontfix (although perhaps an error message
would be helpful.)

[reimar]

This is wontfix, second-guessing the user is not an option to me, and a really
low value for -cache-min works out well enough for me.