SIPCORE IETF-83

Contents

Chair slides

Proxy features

Open issue: Information for IANA registration of feature tags to be used with Feature-Caps.

WebSockets Transport for SIP

Date: Tuesday, March 27 2012, 1710-1810
Location: Paris, France
Chairs: Adam Roach, Paul Kyzivat
Notetaker: Vijay Gurbani

Chair slides

Presenter: Chairs
Slides: http://www.ietf.org/proceedings/83/slides/slides-83-sipcore-2.pdf

Mechanism to indicate proxy capabilities is now a chartered item.

Since IETF-82, two RFCs published.

3265bis on IETF LC.

4244bis headed for 3rd WGLC.

Proxy features

Presenter: Christer Holmberg
Slides: http://www.ietf.org/proceedings/83/slides/slides-83-sipcore-1.pdf

Wants to spend most of the time on mechanism itself.

Hadriel added as author.

Requirements draft heading to WGLC.

Mechanism draft recently (Mar-1-2012) became a WG item.

Open issue: Information for IANA registration of feature tags to be used with Feature-Caps.

Andrew Allen: We should use feature tags with a ref. to a spec. defining behavior.

Discussion ensued between Hadriel and Partha on "entities without a Contact header" and whether or not they feature tags

Longer discussion ensued on topic

RjS: Brought up the issue of company name as feature tags and Keith expanded why naming tags after companies is not a good idea.

Hum for determining if a separate table is needed or use the same table.

Hum 1: Same table as is currently in the registry

Hum 2: Create a new IANA registry.

Consensus: We will create a new IANA registry for Feature-Caps

WebSockets Transport for SIP

Presenter: Victor Pascual
Slides: http://www.ietf.org/proceedings/83/slides/slides-83-sipcore-0.pdf

Went through history. Couple of running implementations and more interest from more.

Main behavior is between websocket UA and first hop in the network. Proxy-to-proxy websocket is for future use.

Sal: proxy-to-proxy websocket will not work anyway.

Problem: How to get requests coming in the backwards direction, i.e., to the ws client.

Solution space:

Option 1: use normative language MUST use outbound.

Option 2: use non-normative, outbound as preferred but other alternatives

Option 3: do nothing

Bernard: Option 1 is problematic. Don't mandate it.

Christer: We need a option 4: REG that has succeeded, and edge proxy can send requests on existing connections.

Seemed the room was happy with Option 2.

Bernard: Have you seen any weird firewall interactions with WebSockets?

Victor: We have not tested extensively.

Sal: Talked to some folks from Google who have deployed WebSockets without problems.

Quick poll and hum indicated WG support for adopting as a milestone. Will be verified on list.