Cullen Jennings' Notes from IETF 63 Informal P2P SIP BOF
There was substantial interest in the topic of P2P SIP and around 125 people showed up. It is clear that work in this area is being done by many people including, universities, start up companies, small and large established vendors, and large service providers.
The group discussed some deployment scenarios that might be able to take advantage of P2P technology including:
- Emergency Ad-Hoc: Disaster and emergencies situations where connectivity to the global internet, including DNS root servers, was not possible
- SOHO - Small offices or home networks where the devices discover the other nearby devices and can perform services across them
- Large Scale: Networks with no central server so that they are very cheap to operate and run but support many users
The group discussed some problems where P2P approaches might be a good solution including:
- Distributed Location Server: Use a distributed P2P system to distribute the register/location server information
- Configuration Storage: Store configuration details for a phone across the networks
- Voice Mail
- Address Book Storage
- Reputation Systems
- Media Relay: (Distributed TURN servers)
- Media Anonymization
Some issues that people would like to work on:
- Scale: Both scaling up and down
- Trust Models
- Data Storage
All the systems described had somewhat common approaches in that they were using DHTs as the underlying approach. Attention should be paid to when it is the right time to use P2P technology and when other approaches such as DNS and Multicast can provide a better solution.
Interesting comment that we should publish some of this before it all gets patented.
The HIP folks has formed both an IRTF and an IETF group. The folks here should probably go look at these to gain a better understanding of the function of both.
Next steps: There was some talk about what it might take to form a IETF working group. There was substantially more talk about forming a IRTF Research Group. There was not much feeling that this overlapped with the interests of the current p2prg. Many people has a strong desire to have a full day metering where we could discuss this work in depth. Having this in Vancouver just before the November IETF meeting seemed like the most realistic possibility and many people said they would come to this if it happened.
Henry called for volunteers to help with this and help manage this, And offered to provide help himself. Everyone provided mutual encouragement to join the email list and send some comments.
Raw Notes Below
David Bryan presented a few slides on what he is working on. Uses a DHT based on sip message to build a distributed a Very many different ways that you could implement this
Work to be done
- use a DHT to do location vs. doing it in SIP
- traversing NATs
- using media relay NATs
- how do you know who is who
- various routing algorithm, which to use
- Small enterprises simple admin
- Emergency - ad hoc comm.
- Large scale global communications
- Identity issues
- How to deal with DHT parameters for this type problem
There is p2psip.org page - community page. Glad to add stuff from everyone to page.
Phillip Mathews - Nimcat
- having small office system
- working on commercial system today
- interested in DHT methods
- robustness important
- wrote a draft at previous IETF
- talks about things that are important
- what we need:
- Interesting problem s: How to scale,
- Their P2P layer is separate from SIP level. Interesting question - Alan J. draft, separate the P2P layer from SIP and doing it SIP. Not 100% clear what is best...
- Also use multicast to discover nodes. Also broadcast presence information via multicast.
- Everyone knows where everyone is via multicast registrations
Henning: 3 ways he has herd of to map
- Cord like system
- Copy of registration everywhere
- Use DNS to map
- Conducting research in P2P
- Want to connect consumer devices using SIP. Many types of works.
- Similar to some of existing work.
- Consumers devices have very intermittent network connectivity
- Concerned about DHT stability in these.
Working P2P SIP. Understanding distribution of services. Use separation of SIP and DHT layer. Looking at IMS style services.
Do ICE on each node
Use DHT to build a registration table
What can we use P2P for
- Rendezvous signaling
- Configuration storage
- Media storage - is this voicemail. Store voicemail for person who's phone is not available. Need to work even when persons phone is not there.
- Privacy of media. Trust model of forwarding information.
- Identity Assertion - Use P2P node to lead credibility to claim ofidentity.
- Gateway - use gateways is home phones. Cheap gateways. Free World Dialup was originally this
- NAT/Media Relay - two sub cases
- traversal - overlay network.
- Work on picking optimal node.
- overlay networks that route around problems to avoid congestion. Not BGP routing but higher layer.
- David Bryan add -> connect farms of proxies (like autonomic computing) self organizing . Completely conventional SIP UA but use P2P in proxies.
- Interconnect of classical systems and P2P systems
- Cullen add -> Anonymize Signalling and Media
- Cullen -> Address Book
Nick Papadoglou Vodaphone looked at P2P for ringbones and voicemail.
Always run into security and trust problems. Also looked at IMS and p2p for service discovery. Looking at where can I find this server or distributed finding out ring tones.
- Trust Modeless
- How do you do identity. Namespace. Stopping identity threat. Trusting who someone is.
- Data Storage
Separation of basic uses and advanced user. Diagnosis and how to debug a P2P system How does administrator deal with it.
It is hard to debug P2P system because messages are all over places.
Jonathan Rosenberg - don't have to do everything P2P..
There is a lot of interest in this. Document this soon before it all gets patented. Henry S. is willing to write up notes to make them understandable for people who were not at the meeting.
Very roughly 125 people
Eunsoo Shim - Proposal to get a SIP P2P WG and start working on charter for BOF.
Henning had proposal of IRTF group.
Gonzales explaining about HIP. There is an IRTF HIP group and a HIP WG. May want to look at this model.
Concern about IRTF - there is no end to research. Progress is very slow.
Need more time to talk about points of commonality. Workshop, interim meeting. Concrete proposal to have 1.5 day workshop around Vancouver. Perhaps some demos of real things.
May be fairly close to experimental agreement -
John from Panasonic - P2p WG is chaired by co-author of JXTA. Do not have impression. - modeling of HIP approach. IRTF is very long term research thing. Spending next few months for BOF that could happen next year makes sense.
Current p2prg is all over the place. We are focusing very specific problems need faster better thing.
Trying to bring carrier in this effort
At the p2prg at IETF 61, it was less than 10 people. It was place to meet and present what they were doing.
Henry ask for some volunteers to spread the talent and get some project management to get this done.
Things to do
- talk to p2p
- consider IRTF list
- consider setting up meeting around Vancouver