Emergency Context Resolution with Internet Technologies WG
Emergency Context Resolution with Internet Technologies WG
(ecrit)
Wednesday, March 9 at 0900-1130
===============================
CHAIRS:
Marc Linsner <marc.linsner@cisco.com>
Hannes Tschofenig <hannes.tschofenig@siemens.com>
SCRIBE:
Spencer Dawkins <spencer@mcsr-labs.org>
AGENDA:
1) Agenda Bashing
(Chairs, 5 min)
http://www.ietf-ecrit.org/IETF62/ecrit.ppt
- Rohan: We have 65 minutes of material in a 150-minute slot.
- Hannes: we have time to talk ...
2) Requirements for Session Initiation Protocol (SIP)-based Emergency
Calls
(H. Schulzrinne, 15 min)
http://www.ietf-ecrit.org/IETF62/IETF62-ecrit-hgs.ppt
http://www.ietf.org/internet-drafts/draft-schulzrinne-sipping-emergency-req-01.txt
http://www.ietf-ecrit.org/cache/draft-schulzrinne-sipping-sos-04.txt
- Idea was to collect core requirements for emergency calling, independent
of technology and national considerations
- Sipping draft has been around for a bit - cutting out trunk replacement
in ECRIT
- Number provisioning
- Identification of emergency calls
- Location determination
- Call routing
- Focus here is protocol-related requirements only, not operational
requirements (which also matter)
- Some requirements protocol-specific
- Steve Norris - Network resiliency? Priority indication?
- Hannes - these may be operational requirements, important but not in
scope for this working group - we aren't doing system design at IETF
- James Polk - Prioritization is specifically out of scope for this
working group
- Not sure how much we need to do here on number provisioning - idea is
that users dial familiar numbers, but we don't use these as routing
identifiers
- Network boundaries may not match identifier boundaries, but this is
typically national-scale, not local network scale issue
- Roaming - need to provide both visited and home numbers
- Support multiple emergency services (more common outside North America)
- Call identification requirements are reasonable but contradictory - need
to figure out what actually fits
- Universal system-visible identifier, local emergency identifiers
- All entities along a path must identify emergency calls
- Richard - how does this work if you have identifier translation along a
path? which is also a requirement
- Requirement is actually for multiple intermediate translations along the
path
- Backwards compatibility - UACs that don't understand identifiers?
Outbound proxies may not understand identiers - assume that home AOR does,
that's what you actually know about and have control over
- Rohan - software clients are still being updated regularly - don't
assume this constraint - but SBCs have a business case of protocol repair
- Jonathan - the problem is that you don't control the things you need to
talk to
- Not worrying about hotel outbound proxies, etc. at this time
- Location requirements - multiple location providers, by reference, end
party, third party. Assuming home provider is what we need to deal with now
- Rohan - distinct providers of SIP services and IP connectivity? if we're
going to support this, call it out explicitly - it's the fastest-growing
form of SIP services
- Michael - additional complexity with 802.11 mobility - need to call this
out explicitly as well - as long as provider can provide location
information, that's what we need to provide the service - we have to go down
to layer two to solve this
- James Polk - large-scale layer-two networks are more problematic - is
this IEEE scope?
- Civic and geographic information are done?
- PIDF-LO is done?
- Michael - large-scale layer-two networks need to be in scope to meet
real regulartory requirements - but we don't need to say how providers know
where the location actually is
- Rohan - resounding consensus from layer two is to provide raw
information and let IETF define actual protocols - we need to define this if
we care about it - but this is GeoPriv material, not ECRIT material - but do
we need to set requirements for GeoPriv here?
- James - we need routing to the right PSAP to be in scope!
- Jon Peterson - there are already drafts circulating in other working
groups - requirements don't need to be done here, but we can discuss this -
seems hard to do requirements without an architecture design that we don't
have yet - there is a broader problem that we decided to slice into solvable
parts, including this WG's charter
- James Winterbottom- assuming that we already have a dependable source of
location
- Brian Rosen - if we don't have someone responsible for the big picture,
we don't know where to go - someone has to have the big picture; if we don't,
who does? - do we need to get something out quickly, or work on an
architecture? Don't cycle on this for two years before we can do something
- Jon - we are building a tool, not a solution, and it would be
presumptious to do that in the IETF
- Brian - we can't build national systems, we can only build international
systems, and there's no venue to do this
- Jon - we are not the UN, we can convince people our tools are useful,
not sure we can do more
- Richard - we are talking requirements, not solutions - let's focus on
requirements before we do anything else
- Rohan - national groups are looking to the IETF - we aren't the UN, but
nations are asking for help -
- * - if we don't decide, it will be decided nationally
- Call routing has basic robustness requirement- UA or outbound proxy or
home proxies or other proxies need to be able to routing, even if other
entities don't understand non-SIP URIs
- Non-controversial requirements for routing
- Possibly controversial - choice of PSAP (campus vs city police),
traceable resolution? for caller? or just for PSAP?
- Steve Norris - Don't understand choice of PSAP (this is a college
requirement, among other places - industrial, security firms) - we're
talking about getting to the "real" PSAP, right?
- Support for mutiple signalling protocols? Most people are worried about
SIP - can others say something on the mailing list so we don't get
blind-sided?
- Media-capability-based routing? TDD, IM
- Language-based routing? James Polk - this is a good idea - we would need
non-SIP directory services that support this - would this work within a PSAP
network for additional routing after the call gets to the PSAP? - Jon
Peterson - if this is low-hanging fruit, fine, we just need to think more
about this
- Patrik - "lookup must be fast" - Sweden requirement is to minimize the
number of reroutings based on guessing the language
- Michael - if you can identify the words, there are other ways to do this
- hints OK, but don't make this an operational requirement
- Caller identification requirements - language identification?
Identification of supported media? We mostly know how to do these today
- Patrik - is call-back in these requirements? No, but it should be
- Call Setup - Authentication override? Mid-call features (outside the
scope of this working group?)?
- SIP is external interface - allow MGCP, Skinny, H.323 within
administrative domains?
- Rohan - stimulus-oriented protocols aren't appropriate because these
devices usually also speak SIP, H.323 etc.
- Steve Norris - we're talking about voice service - that's the
requirement - but we work on protocols, not on services
- Henning - protocol choice drives idenifier discussion
- Is PSAP OK terminology? Steve Norris - ETSI has had problems with PSAP -
this is a function, not a box or a person - PSAP isn't the only problematic
term we're dealing with, need to scrub this
- Patrik - what about receive-side call blocking? What is interaction with
authentication override? Need to at least discuss this - haven't had
contributions on threats - but you'll get one from Sweden soon!
- James Polk - protocol within PSAP doesn't matter. SIP and H.323 are the
only choices for call routing. Focus on SIP? Jon Peterson - how much does
our work here assume SIP at the requirements level? Would any URI work?
- Jonathan Rosenberg - might be dependent on protocol, or not - can we
break the problem into pieces that are mostly NOT dependent on a protocol?
Use a URN instead of a SIP URI?
3) Emergency Call Requirements for IP Telephony Services
In Japan
(M. Kawanishi, 10 min)
http://www.ietf-ecrit.org/IETF62/ecrit62-arai-japan-req.ppt
http://www.ietf.org/internet-drafts/draft-arai-ecrit-japan-req-00.txt
- Requirements from a draft MIC report, focused on VoIP specifically
- Want VoIP emergency services to be equivalent to existing PSTN
environment
- Stu Gallman - how is priority override actually performed today? This
question is actually out of scope - offline?
- James Polk - actually, may be relevant - overloaded proxy, etc. Japan
worries about this, don't ignore it
- Pointing out that PSAP is the only party that can end emergency calls -
includes reversing call from PSAP
- Pointing out requirement for redirection to altenative PSAP if original
PSAP is unavailable
- Calling-line ID must be presented unless an override code is dialed
4) NENA Requirements for Emergency Call processing
(B. Rosen, 10 min)
http://www.ietf-ecrit.org/IETF62/IETF61-NENA-Requirements-rosen.ppt
http://www.ietf.org/internet-drafts/draft-rosen-nena-ecrit-requirements-00.txt
- Requirements from NENA VoIP/Packet technical committee
- 6134 PSAPs with irregular boundaries in North America - routing based on
side of streets or worse
- Lots of effort to keep civic addresses accurate
- Current system has good accountability for every entity along the path -
matters when something goes wrong, because you debug problems in
life-threatening situations
- Need to know what happened with signalling
- Got to be able to call 911, although other numbers could work
- Need to have calls go to H.164 numbers if you aren't served by 911
- James Polk - is signalling trace like History-Info? Yeah, and Vias -
don't drop Vias at SBCs, for instance
- We assume soft phones get location and provide location on incoming
calls - what if the PSTN comes into a gateway - what's the way to make those
scenarios work?
- Need congestion controls everywhere - don't avalanche problems
downstream elements - congestion control in the SIP sense, and this is at
the endpoint level, not at the intermediate node level
- Multiple locations in a call? Gotta route based on one location - know
tower location plus measured location - this does happen today, and is
legitimate
- Location provider may not be communication service provider
- Need defaults for location when information is missing - choosing "somewhere"
- Eric Rescola - what is scope of default route? there is a hierarchy of
defaults
- Rohan - 911 calls on cell phones in California go to Sacramento
- Steve Norris - can't hold up emergency calls for location information
- * - what happens if you provide coordinates instead of street addresses?
Hours of discussion on this, but existing service uses both and doesn't do
conversion well - what about going forward with GPS-capable devices?
- Address hierarchies - tenants at a location, etc. URI in the database
would be enough
- Additional information about a person - URI would be enough
- Validate before you call - location exists in the database before you
use it to route. Validation databases are also multiple and often map to
PSAP boundaries in North America
- James Polk - who is "you"? - doesn't matter - UA must validate? Or
location information system - place burden on the location server to
validate? doesn't matter, still needs to be validated - UA can validate, but
location service must validate
- Hannes - need to be very precise about what we're trying to accomplish
here - we're skating along the edge of the charter. Additionally, the term
validation is used with two different meanings.
- Henning - we have at least two mental pictures of validation - not just
a terminology problem, but confusion - does an address exist, and am I
actually at that location - how do I know a street address is correct? But
this is clearly out of scope
- Postal address isn't actual community name - need to know which one
we've got
- Some PSAP boundaries are incredibly irregular - river course, all of a
city except for two streets
- Lots of entities will route - routing data must be secure, reliable in
the face of massive failure
- Callback identifier must be reliable
- "No hangup" mechanism - no BYEs!
- Distribute route failure information and related diagnostic information
- Want one requirements document quick - either this one or a design team
draft
5) ECRIT Basic Reqs
(R. Stastny, 10 min)
http://www.ietf-ecrit.org/IETF62/IETF62-ECRIT-basic-reqs-Stastny.ppt
http://www.ietf.org/internet-drafts/draft-stastny-ecrit-requirements-00.txt
- Draft intention was to step back to high level
- Updating based on list discussion
- Don't want operators to withhold location information (even because they
want to sell it)
- Must verify path to PSAP/ECC periodically
- Steve Norris - get location information from user at last resort? Mobile
phones are doing this by voice now in Europe, thinking about better
mechanisms
- James Polk - 802.11 phones in a home won't have their own locations;
$500 boxes may ask users to provide information (although this won't be
reliable) - this is a policy question
- Communication established by external events (is this out of scope?
certainly a user interface issue)
- Indication whether you are allowed to make an emergency call (no
location, no authentication, etc.)
- Have added home and visited emergency URI requirement
- Have added recognition of emergency calls by intermediate devices
- Steve Norris - 3GPP also has requirements for downloading roamed-to
emergency numbers automatically
- Jon Peterson - being able to designate a gateway for emergency calls is
actually in scope, but going beyond this is probably out of scope
- Some SHOULDs are MUSTs in some countries
6) ECRIT Location Scope Requirements
(J. Winterbottom, 10 min)
http://www.ietf-ecrit.org/IETF62/Ecrit_LocationRequirements-winterbottom.ppt
http://www.ietf.org/internet-drafts/draft-winterbottom-ecrit-location-scope-req-00.txt
- VoIP needs the same assurances for PSAPs that they have from POTS
information today
- Location tied to a device, not to a user, at the time that a call is
made
- Location needs to be a real place - postal vs jurisdictional location
- Patrik - location needs to be what a fire truck is using, whether it is
right or wrong
- Brian - same for information that's used to route
- Need accountability for location providers
- Reliability may not be in scope for Ecrit, but it needs to be in scope
for somebody
- Steve Norris - if you have two pieces of location, both of them are
required
- James Polk - Access provider is probably actually providing location,
not service provider
- Jon Peterson - we were talking about scope when we did charter - how is
the protocol depending on the validation of location information, if we
understand that error conditions are included in the protocol?
- Henning - if there's a one-step resolution protocol, this is probably OK
- if it's not one step, if multiple steps of resolution are required and
visible outside the PSAP, the protocol needs to know about validity
- Brian Rosen - it is in scope to require location validation before you
route emergency calls based on location - why is this controversial? It's a
North American requirement, for sure
- Henning - Imagine an MSAG that wouldn't route to a real PSAP - confirms
that a location is valid, but you can't tell that a PSAP assignment is
reasonable or not. Location is more complicated than just "does it exist",
and these requirements are stronger.
- <>Brian Rosen - Patrik was right - what people care about is what Patrik
said; what people care about is whether locations work for dispatching fire
trucks - location will be used for routing AND given to the PSAP for PSAP
response</>
7) Next Steps and AOB
(Chairs, 5 min)
- Hannes: We have seen a number of drafts about requirements, some more
mature than others. We need to come up with requirements really soon.
Henning's document is the most complete one (and available for a long time
already). It needs to be updated and have terminology scrubbed. Can other
document authors contribute to one document? Can we pick an editor?
- Brian Rosen - this is actually more collaborative, but I agree
- Roger Marshall and Brian Rosen volunteered to be editors - Roger will
take the lead here
- Rohan - who are these requirements for? All of the drafts have
requirements on all sorts of entities - need to organize the requirements by
who they reply to.
- Rohan - security analysis - need to do much more work in this area.
- Steve Norris - NGN TiSPAN group has IMS requirements that are 90-percent
overlap with 3GPP. Emergency services have requirements, too (provider call
release, etc.)
|