Managing, Ordering, Distributing, Exposing, & Registering telephone
Numbers WG
IETF 93 Tuesday Afternoon session II - 15:20-17:20
15:20-15:25 (5 min) Agenda Bashing Chairs
Note Well
Agenda Bashing
Working Group Milestone Review
Date Milestone
Mar 2016 Submit Architecture Overview Draft to IESG (Informational)
Jul 2016 Submit Information Model Draft to IESG (Standards Track)
Jan 2017 Submit TN Resolution Draft to IESG (Standards Track)
Apr 2017 Submit Enrollment Data Access Process Draft to IESG (Standards Track)
Jun 2017 Submit Enrollment Process Draft to IESG (Standards Track)
15:25-16:15 (50 min) draft-peterson-modern-problems-01 Tom McGarry
Modern Problems
- problem statements – nothing new
- actors – some new
- mechanisms – not much new
- use cases – all new
- distributed registries and data stores – all new
Problem statement review
Actors – new in 01
- Separation of num auth from registry (focus on work of registry)
- Service enablers
- Should there be two categories of actors
o Adminatration
o Service management
- Assignee and Delegate
Henning – two types of actors
- logical actors such as regulators
- someone who actually participates in the protocol actions
- consumers might have access to a subset of information about their number – need to be clear if this is included as an actor
Tom – delegate is reseller
Henning – delegate is someone who canŐt get access to protocol directly but can have some access (write access) to profile
Jon – Need to separate actors that participate in the protocol
Assignee and delegate are roles
Chris – can it be defined in the context of who is making the http request?
Jon – assignee has a business relationship
Henning – CSP and PBX context, small business could use the same protocol as CSP
Worried that the model isnŐt general
Jon – Nothing restricting making it general
Chris – are consumer and csp the same?
Henning – could use the same bits on the network. Consumer to csp, csp to registrar but requests could be the same. Names should reflect roles in numbering management
Chris – need to discover if the interface for the end user is the same as interface for the CSP
Tom – user has different requirements then csp such as amount of inventory
Jon – use cases will expose this
Martin – agree with henning, keeping it at level of assignee/delegate with relationship, keeping policy out of discussion, protocol need to identify information conveyed, policy will determine what information is sent between players
Henning – for use cases it is ok to informally use terms, not ready to make them formal protocol roles
Mechanisms
- acquisition
- management
- retrieval
Use Cases
Acquiring TNs
- CSP acquires TN
o Is communication of qualifying info part of the protocol
Jon – modern protocols are not going to give utilization forecasts, will be some out of band information – probably not part of the protocol
- CSP requests TN
Henning – what is the minimal set of functionality we need to support
Tom – interconnect agreements part of problem
Chris – does inventory batch operations need to be part of this?
Tom – part of the way it works today
Jon – likely to need to include batch operations
- Registry assigns TN to CSP
Use case – Acquing TN
- User/Delegate acquires tns from CSP
- Very similar process
Pete – some of it is internal and not protocol, hard to understand which is protocol and which is not, helpful to separate
Martin – existing model requestor is currently carrier, govt authority has a degree of control, end users and corporations have little information, those users have relationship with carriers today and carrier has the information, more information will be needed for non traditional roles
Tom – profile date, right?
Jon – to petes point, credentials requirements is clearly protocol, problem statement is not specification, to martins point, enrollment is where martinŐs point is true
Andrew Allen – request procedure – will it take into account vanity numbers? It needs to
Tom – need to specify what the requestor is wants, specific number, range, any, etc.
Chris – thinks it is needed
Henning – will be search criteria, for instance area code all the way to a specific number. Listing numbers that are available will result in millions of numbers
Allen – more of a search string like a domain name
Andry? – already have ability to search on area code
Use case – accessing numbering information
- Two types of information (administrative, service
- Info can be stored at registry or CSP
Use case – accessing numbering information
- service information access
- Privileged access for govt entities
Use case – service management for numbers
- service management
- Updating service information
- Updating administration information
- Changing CSP
- Terminating service
Distributed registries and data stores
Henning – based on prototype, multiple registries might be needed for redundancy purposes. Assuming that sync between registries are worked out. Actors are internally consistency in what they do. So donŐt need
Chris – agree with henning
Jon – be good to have text from experience from prototype into the draft
Daryl? – wish this happened 7 years ago (speerment, drinks, etc). calling name part of this? No longer limited by SS7 15 digit limitation.
Jon – proposal for WG for CNIT. Profile will surely have calling name accessor but there are political considerations
Henning – how do you access version how do you validate. Validation is not in scope of modern for calling name.
Next steps
16:15-16:35 (20 min) MODERN Architecture deliverable Jon Peterson
Moving Parts
- acquisition protocol
- management protocol
- query protocol
- these protocols access overlapping data
- surely this is a common information model?
Telephone-Related Information (TeRI)
Jon – does this make sense?
Henning – at abstract level yes, maybe not at bits and bites level, in the end it is a database record. Two types of records so that not all information needs to be stored everywhere. Assume there is a telephone record that might have pointers to other records. We donŐt worry about how other pieces are managed, just have pointer to it. Would be helpful to recognize that not all data needs to be replicated and that there will be indirection to things like OCN records.
Chris – word management might be too strong, records should be simple,
Jon – name might change, provisioning?
Chris – is admin info and service info really different things?
Jon – some of this is based on desire to reuse things like done in weirds
Chris – personal preference to keep things simple
Mapping the Modal to an Instance
- TeRi records would live in servers
- All sorts of entities might manage or query
Jon – question of how to organize the work
Andy Newton – are end users always related to telephone system. Is law enforcement one of the queriers of info
Henning – some numbers might be public
Andy – now understand why wierds is being considered
Operations and Records
- proposal – define all three protocols in terms of this TeRI model
- each protocol will have its own operations but operate on a common class of TeRI records
- Operations (query, response, etc) will have their own source, subject and attributes
- TeRI records contain info about TNs
TeRI Record Contents
Transport and Encoding
Chris – the words Ňdata modelÓ are important here. JSON makes a lot of sense. What do we think the data model will look like, how extensible, pretty simple?
Jon – crawl before we work. Look at semantics first, then worry about transport
Henning – will have small set of universal data objects. need to consider IANA considerations, no x-, p-
Dan York – digital signatures?
Jon – some level of digital sigs in objects
Chris – follow CRUD model
Jon – this is the easy part
Peter – will all these protocol run on public internet
Jon – assume they can run on both public and private networks.
Peter – no requirement that it be run only on private internet? Operation and security consideration might be different
Henning – separate issues – name space separation for instance military networks. Enterprises have ability to allocate numbers
Jon – no reason to disallow
Dan York – need to think about this as accessible over public internet. Is intent to write requirements draft?
The TeRI Suite
- Core TeRI model
- TeRA
- TeRM
- TeRQ
-
Jon – currently on the hook for three protocols. Proposal is to extract out the TeRI model. Is this reasonable and how do we execute.
Chris – have a single data model
Jon – should we change charter to reflect this
Henning – keep it simple as possible, distinction between acquisition and management is artificial. Single document might be enough.
Sanjay – are you saying to separate out common element
Jon – separate out core data model first then understand protocols
Steve – Arch spec might be good place for this
Cullen – doesnŐt matter how many documents
Martin – agree with Henning and Chris in that they can all be worked at once, keep it simple
Allisa – charter is structured to allow this, milestones can change.
Andy – if we are doing multiple protocols then need data modeling
Jon – data model semantics first
Chris – established ways to handle this, potential approach to use requirements to define actions
Jon – need to avoid making requirements to big
Daryl – requirements docs can sink the process. There are RFCs that can be mined.
Jon – Charter compels us to do that.
Daryl – a lot of work has been done, lets leverage it.
16:35-16:55 (20 min) draft-peterson-terq-04 Jon Peterson
Why present TeRQ today?
The TeRQ Architecture
TeRQ Operations
??? – what is the query source, user or box?
Jon – could be domain name, or URI
TeRM Operations (hypothetical)
TeRQ Base Element Types
TeRQ Profiles
Next steps
Chris – http post put get with conventional mechanisms to map to crud operations
Jon – No re-invention at that level, focusing on semantics
Chris – can define that in data model
Alyssa – Do you envision that the intermediary can perform all operations
Jon – envisioned intermediary as a proxy
Alissa – intermediary might add some complexity
Jon – Query source and query intermediary defined in TeRQ
Chris – Intermediary could be doing other things than just proxy
Jon – Yes
Henning – unconvinced we need to specify the processing of the intermediary
16:55-17:20 (25 min) Modern Prototyping Henning Schulzrinne
Overview
Architecture
Authorization Model
Operations needed
Paxos (& similar dist. Consensus) assumptions
Paxos for distributed consensus
Paxos
Prototype
Implementation
Andy Newton – RDAP component, did you implement
Henning – in progress, need input on if or how to implement
Andy – See a benefit as clients for other registries can easily interface with this registry
Henning – Exactly why its being considered
Chris – with PIN, implying a secure token mech
Henning – yes, inspired by DNS registry model, Some notifications required for instance in number portability
Jon – seems like awesome stuff, where do you think it fits in?
Henning – lead into protocol spec