Notes On OMA Ad-Hoc at IETF 62
Produced by Spencer Dawkins
Starting out with a pretty packed Rochester Room (and food
delivered to the previous meeting), after Dean found the projector...
Dean is current IETF liaison for the OMA
Cisco providing dinner...
Improving collaboration, informing IETF, discussing specific OMA
activities with related drafts
RFC 3975 defines our relationship with OMA
- OMA SIP/SIMPLE Presence composition and attribute issues/requirements
compared to IETF (SIMPLE data model, RPID etc.), future requirements -
• Krisztian Kiss presented OMA
Presence 1.0 release
• OMA has done work on presence
composition policy
• Unique identifier for an
OMA-defined service (POC service), with service descriptions
• Presence server resolves
ambiguity in OMA-defined attributes
• Default composition policy is
"choose latest timestamp", but other policies are possible
• OMA view - presence server
should not leave ambiguities for the watcher - determine actions based
on source of presence information?
• Makes watcher simpler, but
presence server is less extensible and needs knowledge of semantics and
sources
• Presence attributes -
Person/device/tuple model
• Have added willingness,
application-specific availability, overriding willingness, along with a
lot of reused attributes from IETF
• Network-availability,
Session-participation are also new from OMA
• PIDF dependencies - using RPID
parts, may use rest of RPID, PRESCAPS, CIPID, FUTURE, etc. in later
releases
• Using service description
instead of PRESCAPS? Yes, for POC, but leaving the door open for other
services
• Device-ID will be 128-bits and
random - OMAI needs to be registered with IANA
• Authorization issues -
extensions for <transformations> - may have requirements for
prescaps, cipid, future, XCAP hard-state publication - will IETF define
authorization rules for these I-Ds?
• Added <other-identity> as
default policy - overlap with <any> attribute? only works if
there is no other rule matching - this may be going against current
policy model
- OMA SIP/SIMPLE IM requirements (for the release targeted for 09/2005)
- Markus Osomaki
• OMA Instant Messageing 1.0 -
not as advanced as presence work at OMA, target date is toward yearend,
can still use stuff IETF is defining now
• Requirements to IETF coming
more from OMA and less from 3GPP?
• OMA IM based on SIMPLE work,
with strong inclination to keep system components compatible with IETF
protocol specifications
• At the same time, there is
strong industry demand for finalized system standard in timely fashion
• SIP MESSAGE must be
"store-and-forwardable" when recipient is offline, with notification of
later, final delivery
• MSRP for a single message to
overcome size limits in SIP MESSAGE (need to say "one message" when you
set up MSRP session) - got plans? Sendonly? We have plans, we just need
conventions - are you sure you want this? It's not what Jonathan's cell
phone does today - could be done at receiver UA? This is moving the
pager-mode/session mode problem into MSRP - do we want to go here? List
discussion would be good
• Aki - problem here is that we
have two paradigms and no indication when you use which paradigm -
don't want to pop up "incoming session" for one message - SIMPLE needs
to solve the transition problem that has been identified (here and
previously)
• Is this for page mode or both
modes? Mostly page modes. Doing anything with consent? That's a good
question - if you sign up for OMA service, you're signing up for this,
too.
• Relationship between
store-and-foreward and MMS? Why two mechanisms? No overall messaging
framework - this isn't a new question. MMS isn't available everywhere,
may have SIP-only devices.
• Need message delivery lists
(based on SIPPING URI LIST draft)
• Need private messages within
multiparty (as in SIMPLE discussion)
• Need nickname identity within
chat room (as in SIMPLE discussion)
• Data manipulation requirements
- OMA is working on XCAP-based mechanism - augment with WebDAV?
• Authorization for resource lists
• vCARD network-stored phonebooks
as extension to resource lists
• Search PoC and chat conferences
within a scope based on certain attributes - owner, start time, etc.
Also search for people - interdomain? Not clear, requirements focused
on one domain now
- OMA PoC issues
- draft-allen
• Tom Hiller presenting
• POC documents are available
outside OMA
• POCv1 is focused on audio
• Uses SIP, and builds on IMS,
but doesn't preclude non-IMS SIP networks
• Two kinds of SIP servers, the
focus and the application server
• Participating PoC function and
controlling PoC function (acts as focus, media mixer, etc.)
• Two call models - on-demand
calls (INVITE-initiated) and pre-established sessions (fewer bytes and
maybe no SIP signalling on terminating endpoints - uses REFER instead
of INVITE - more state in PoC servers, and maybe proxies, too)
• CDMA has 8000 bps in reverse
direction - that's the pain level we're dealing with
• Latency challenge - slow links
with dormant/activation delays ("under two seconds" requirement)
• Does PoC server maintain
ordering? When two people PoC to me (glare)? PoC may pick one, may
buffer, etc.
• 183 Session Progress with
P-Answer-State: Unconfirmed - allows talker to talk while network is
setting up channels, etc.
• Does PoC server transcode,
since you haven't negotiated media? Yes
• There is clipping (PoC server
sends to all receivers when first receiver answers) - goal is to put
buffering in one place in the network - users can be trained to ask
"are you guys out there?"
• What IP address are you sending
to? the buffer inserts itself into the path. Modified SDP? No, INVITEs
go to focus, anyway - is this a legal RTP transmitter? do we get
problems with RTCP, etc? What does the spec say?
• How are you doing floor
control? There's an OMA-specified floor control protocol - haven't
ruled out BFCP, but need to publish now, so...
• P-Alerting-Mode - only current
value is Manual Answer Override, but manual and auto are added for
future use
• P-Answer-State - allows a PoC
server to tell a talker how far the speech path has been established
• Why not preconditions? Was
considered, design discussion is in the specifications - there are
technical concerns, but a lot of phobia around IPR on preconditions
• P-Alerting-Mode sits in REFER,
and should be Refer-To - this is just wrong. Can it be fixed? Don't
know what resistence is within OMA, people are building in parallel
• Is auto-answer also used for
MSRP? Don't think this has been discussed in OMA - MSRP was set up in
INVITE/200 OK, maybe OMA should think about this? How does this
interact with early media? Is a lost early-media message the same as
lost speech? ("Have you ever talked to your children?")
• What is authentication policy
for Manual Answer Override (so not everyone can just start talking out
your phone)?
• P-Answer-State is
Confirmed/Unconfirmed
• Jonathan - we haven't solved
unconfirmed problem in general SIP space - how to break in on a caller
talking to my voice mail? Almost like shared-line-appearance, voice
mail server is also a focus - this is exactly the same problem in OMA -
mechanism is the same except voice mail is immediate - can't rely on
anything by the phone (in a high-latency mobile environment) -
terminating PoC server becomes a conference bridge - problem at OMA is
that all the buffering is happening at another PoC server, so no
buffering happens at the terminating PoC server
• Is this the simulcast problem?
We need to solve this anyway (so we can synchronize playout)
• Why is NOTIFY used? Jonathan is
modeling this as a shared line appearance (Dean's answer: "because you
don't want to do an SDP media exchange on every call setup") - latency
to PoC client is so high that we can't use JOIN
• Aki - agree with Jonathan that
his approach would work, but OMA also cares about minimizing the number
of messages - three more messages = "bad"
• Jonathan - probably nothing
wrong, why isn't this an isfocus parameter in 200 OK?
• Steve Norris - if PoC server
anticpates what is happening, what happens if it doesn't happen? You
just get a "bong" later on - call failed
• Jonathan - another way to model
this is as a conference out-dial?
• Steve Norris - why is media
buffering before the final PoC server? Just to keep all the buffering
in one place - there can be an arbitrary number of PoC servers inline,
so trying to minimize buffering all over the network
• P-Alerting-Mode used to be in
Callerprefs and disappeared for a good reason, but it's come back as
part of MMUSIC loopback diagnostic testing and simulcast application -
hard problem is always authorization, and whitelist authorization isn't
bad - this really has more general application, it's not a P-header any
more and should be generally defined
• People are going to use this
P-header for autoanswer everywhere anyway when it becomes available -
we should do it as a standards-track specification - to the SIP working
group? But we need this already - can the draft be in on Monday?
- Future issues - ?
• What are OMA interfaces arouund
location roaming and Privacy Profile Register? Do we need to coordinate
with OMA and IETF? It's going into PoC 2 now