[vlc-devel] Re: kasenna ff/rw streaming so close now...
Måns Rullgård
mru at inprovide.com
Fri Oct 21 03:21:02 CEST 2005
Glen Gray <glen at lincor.com> writes:
> On 20 Oct 2005, at 22:52, Måns Rullgård wrote:
>> Maybe it's irrelevant, but the Kasenna server often does crazy things
>> when switching to from trickplay modes. On one occasion, I had it
>> jump forward several minutes when starting ff, and then jump back the
>> same amount when resuming normal play.
>
> Well, it's probably doing what you told it to do :). Remember, the FF
> stream is 1/12th of the normal mode stream. So if you pause and then
> issue the play with scale value 12 for the FF stream, and use the npt
> start from the same time you paused at in the normal stream, you're
> start point in the FF stream with well off (by a factor of 12).
> You'd need to basically use the pause time and divide it by 12 to
> start the FF stream at the approximate location of your normal
> stream. Or at least that's my guesstimate of what I've ahead of me
> once I get the stream decoded. Fun indeed.
It was no fault on my part. I fed the machine with a single video
stream, and it created the trickplay streams all by itself. Only one
video had an offset of several minutes. Mostly it was around 30
seconds, sometimes even acceptably low. The stream switching should
all be handled invisibly by the server in response to ff/rw rtsp
commands.
This was all after I managed to convince it accept the files at all,
rather than spit them out, telling me the bitrate was too low, which
was obviously not the case. I have zero confidence in Kasenna
products, having had nothing but trouble with them.
> The problem I have is to do with the decoding of the MPEG stream.
> It's a different stream so I need to flush something to get it to re-
> setup the stream params. What that is I don't know yet. I'd hoped
> setting the i_pcr and i_pcr_start to 0 would be what I need, but
> obviously not.
Well, VLC tends to be very particular about the stream parameters.
I'm afraid I can't help you there, though, since I know next to
nothing about the internal workings of VLC.
--
Måns Rullgård
mru at inprovide.com
--
This is the vlc-devel mailing-list, see http://www.videolan.org/vlc/
To unsubscribe, please read http://developers.videolan.org/lists.html
More information about the vlc-devel
mailing list