[vlc-devel] Re: Sgi mediabase and XDMA streams

Derk-Jan Hartman hartman at videolan.org
Wed Mar 24 14:51:23 CET 2004

K, kasenna and mediabase really sucks.
Small overview of what i found out so far.

1: They use different RTSP implementations for their MPEG1/2 and their  
MPEG4 streams
2: MPEG1/2 rtsp does not do sdp's but rtsl or something
3: some of the sessions don't like the SETUP live.com requests.. TODO:  
find out what the servers normally returns after a SETUP
4: They use mediabase 'metafiles'
5: The multicast UDP streams from kasenna are probably meant for rtsp?  
not valid TS

BTW. working with these highbandwith streams over long distances, i'm  
also getting increasingly more aware of the possible benefits of packet  
reordering :)

Example of 1:
Works with vlc.
Though the stream behaves really weird. it gets out of sync after a  
couple of secs, reports PTS errors does several Late picture skipped  
messages and gets back on track again (infinite loop of this  
This may be of interest?
[00000225] livedotcom input debug:  
[00000225] livedotcom input debug:  

Example of case 2:
I understand this x-rtsp-mh was scrapped from the standard?

Sending request: DESCRIBE rtsp://videoserver.nob.nl/demo/r_kelly  
CSeq: 1
Accept: application/rtsl
User-Agent: VLC Media Player (LIVE.COM Streaming Media v2004.02.26)

Received DESCRIBE response: RTSP/1.0 200 OK
CSeq: 1
Date: Wed, 24 Mar 2004 12:54:05 GMT
Server: Kasenna MediaBase version 6.1.2
Content-type: application/x-rtsp-mh
Content-Length: 412

Need to read 412 extra bytes
Read 412 extra bytes: <MediaDescription>
NumStreams = 1
StreamID = 0
MimeType = application/x-sgimb
MaxBitRate = 6486621
Height = 576
Width = 720
FramesPerSecond = 25.000000
AspectRatio = 0.673500
FramesToIFramesRatio = 12.000000
AudioHeader = 0
Duration = 265802029
StartTime = 0
PlayModes = 1
TypeSpecificData = MPEG-2
VideoPid = 17
AudioPid = 18

I might try to patch limited support for this into liveMedia.

Example of case 3:

Example of 4:
return the following file:


Example of 5:
It's TS without PMT. so essentially barebone ES's in an oversimplified  
TS stream. This might work when an RTSP session is controlling the  
stream, but it isn't valid TS and therefore VLC doesn't like it.
VLC might be able to do a sort of autospawn of decoders on each newly  
discovered PID, but it's a workaround at best (though probably what  
other players like STB's and Optibase player do, since they do play  
this stream).


On 05 mrt 2004, at 01:25, Ross Finlayson wrote:

>> The windows 0.7.1 version actually got further (newer live.com  
>> build). However i couldn't copy the debug messages live printed in  
>> the windows-console. It found a audio and a video stream, then  
>> stopping with:
>> livedotcom debug: RTP subsession 'video/MPV'
>> livedotcom debug: RTP subsession 'audio/MPA'
>> livedotcom error: PLAY failed No RTSP session is currently in progress
>> The part it doesn't like seems to be:
>> after Sending request: SETUP
>> It shows:
>> Transport: RTP/AVP;unicast;client_port=2310-2311
>> adn then
>> Received SETUP response: RTSP/1.0 461 Unsupported Transport
>> Message: Unsupported Transport for this format
>> (Which is actually the same thing QT reports when you try this.)
> That's the problem: the RTSP server says that - for some strange  
> reason - it can't do standard RTP unicast streaming.  I.e., the  
> problem is with your RTSP server.
> When I try opening
> rtsp://popeye.surfnet.nl/production/rug/rug_stadtman/ 
> hartvannederland_130204.mpg
> the server reports:
>         Server: Kasenna MediaBase version 6.1
> So, you should ask "Kasenna" why their product is non-compliant.
>> I tried xdma (and rtsp version of the link) in QT, RealOne and  
>> mplayer and none of them understood it.
>> So I was wondering how far from rtsp this format is, and whether  
>> perhaps live.com could have a compatibility mode that would support  
>> it?
>> How feasible is this?
> Not very.  I'm generally disinclined to support non-standard protocols  
> and formats (although if Kasenna, or anyone else, is interested in  
> paying my company to do this, I might consider it :-)

Universiteit Twente
Derk-Jan Hartman (d.hartman at student.utwente dot nl)

This is the vlc-devel mailing-list, see http://www.videolan.org/vlc/
To unsubscribe, please read http://developers.videolan.org/lists.html
If you are in trouble, please contact <postmaster at videolan.org>

More information about the vlc-devel mailing list