[vlc-devel] VLC RTSP Server Support for Keep-Alive Messages to Kill Orphaned Streams?

[Robert Krakora]

2009/7/15 Rémi Denis-Courmont <remi at remlab.net>:
> Le mercredi 15 juillet 2009 20:27:43 jpd at videolan.org, vous avez écrit :
>> On Wed, Jul 15, 2009 at 12:59:17PM -0400, Robert Krakora wrote:
>> > I would be glad to work on this.  I have a lot of experience with a
>> > variety of RTSP server and RTSP client stacks.  Usually, clients seem
>> > to send benign RTSP requests such as GET_PARAMETER to keep the session
>> > open.
>>
>> It seems to me that RFC2326 10.8 explicitly says GET_PARAMETER without
>> option may be used ``to test client or server liveness'', or when
>> already playing, 10.5 says PLAY without range may be used by the client.
>>
>> Presumably failing enough liveness tests would permit one to consider
>> the session dead and release the associated resources.
>
> Guys, this has already been discussed in this mailing list. and I fail to see
> any new element that would grant re-opening the topic.
>
> The point is, because RFC2326 is under-specified. The server MUST support
> every valid ways for a client to ping. Otherwise, it can time out live client.
> IMHO, that would be worse than not timing out dead clients.
>
> Handling GET_PARAMETER and/or PLAY would fail.
>
> --
> Rémi Denis-Courmont
> http://www.remlab.net/
>
> _______________________________________________
> vlc-devel mailing list
> To unsubscribe or modify your subscription options:
> http://mailman.videolan.org/listinfo/vlc-devel
>
>

Guys:

Again, thanks for the quick response.  May I inquire how then does the
VLC RTSP server handle orphaned RTSP sessions?  All of the proprietary
RTSP servers (VOD) with which I have worked utilized both
GET_PARAMETER and PLAY to determine if a client is still "alive".
This establishes a constraint that the to which the client must
adhere.  I realize the VLC dilemma in that it has a much broader user
base that the proprietary RTSP servers and, therefor, must maintain
interoperability with a broader range of RTSP client stacks.  How
about a command line and GUI option that allows the minimum
"keep-alive" functionality described in the RFC to be utilized (i.e.
GET_PARAMETER and PLAY).  Of course the RTSP server would also have to
convey the "keep-alive" timeout to the RTSP clients so that they know
to send "keep-alives" and at what interval to send them.  Surprise
removal of RTSP clients without TEARDOWNs being issued happens all the
time with settop boxes in situations where they lose power (unplugged
for instance) or are reset by users.

Best Regards,

-- 
Rob Krakora
Senior Software Engineer
MessageNet Systems
101 East Carmel Dr. Suite 105
Carmel, IN 46032
(317)566-1677 Ext. 206
(317)663-0808 Fax