RE: draft-ietf-mmusic-sip-session-timer-00.txt

["Donovan, Steven R. (MCI)" <[e-mail]>]

> -----Original Message-----
> From:	[e-mail] [SMTP:[e-mail]]
> Sent:	Thursday, February 11, 1999 1:02 PM
> To:	Donovan, Steven R. (MCI); '[e-mail]';
> '[e-mail]'
> Subject:	RE: draft-ietf-mmusic-sip-session-timer-00.txt
> 
> Pardon the perhaps very naive question, but I suppose I am not as versant
> in 
> SIP as I would like to be.
> 
> My understanding is that a SIP Client issues an Invite message to a SIP
> Server,
> which in turn forwards the request to a target SIP Client (or some GW or
> some
> sort). Once the ACK has returned to the originating client, I was under
> the
> impression that both SIP Clients send RTP packets directly to each other,
> bypassing the SIP Server (this is where I am unclear).
> 
	This depends on the implementation of the SIP Server.  There are
valid reasons to have the SIP Server proxy the IP packets.  But in general,
you are correct.  There will be SIP Servers that do not have a view of the
RTP flow.

> So if this is true, how will a SIP Server (stateful, or stateless) enforce
> any
> session timeouts? A malicious SIP Client can simply keep the session
> active
> for as long as it wishes, ignoring any messages from the SIP Server that
> may
> tell it to disconnect.
> 
> This, as far as I can see, could cause some serious problems if one wishes
> to
> deploy a service and still maintain some control over the sessions.
> 
> Again, perhaps I am way off here. Any clarifications would certainly be 
> appreciated.
> 
	You are correct and I personally would not rely on reporting from
the SIP Server to do reliable billing.  However, if a stateful SIP Proxy
server is built, for whatever reason, the current SIP specification does not
have a mechanism for the SIP server to know which of the sessions for which
it has state are valid.  This draft gives the SIP Server a mechanism to
remove those sessions that it feels are no longer valid.

	The amount of control that a SIP server exerts over sessions is a
separate issue.  As John Hearty accurately points out in a separate email,
the user of the SIP device can choose to bypass the SIP server entirely.
The challenge for service providers will be to implement services that make
user's want to access the service provider's SIP server.

> PatC
> 
> >This time with the draft attached.  Sorry for the extra mail.
> >
> >Steve
> >
> > <<draft-ietf-mmusic-sip-session-timer-00.txt>> 
> >
> >> -----Original Message-----
> >> From:	Donovan, Steven R. (MCI) 
> >> Sent:	Thursday, February 11, 1999 10:17 AM
> >> To:	'[e-mail]'; '[e-mail]'
> >> Subject:	draft-ietf-mmusic-sip-session-timer-00.txt
> >> 
> >> This message is notification of the following Internet Draft which has
> >> been submitted to the IETF for posting.
> >> 
> >> <draft-ietf-mmusic-sip-session-timer-00.txt> 
> >> 
> >> Abstract: 
> >> 
> >> This document proposes an extension to the SIP specification.  This
> >> extension adds a new message header that is used to specify the 
> >> duration of a requested session.
> >> 
> >> The session timer can be used to limit the total duration of a session
> >> if, for instance, one of the participants in the session wants to
> >> limit the cost of the session.  It can also be used by stateful SIP
> >> Proxy Servers to track the status of sessions for which session state
> >> exists on the servers. Currently a stateful SIP Proxy Server that is 
> >> not handling the media stream(s) for the session has no mechanism to 
> >> definitively determine the state of all sessions for which it has
> state.
> >> While the SIP Specification does provide the BYE method for terminating
> >> the session, there is no mechanism for a SIP Proxy Server to detect the
> >> end of a session when the BYE message is not sent or is lost due to 
> >> network problems.
> >> 
> >> The internet draft is attached.
> >> 
> >> Steve Donovan
> >> MCIWorldcom
>