P2PSIP notes, IETF 78
17:40 Monday, July 26, 2010. Brussels Conference Room
Notes by Dale Worley

----------------------------------------------------------------------

Cullen Jennings presents draft-ietf-p2psip-base-09:
REsource LOcation And Discovery (RELOAD) Base Protocol

slide:  Overlay Link Protocols

Question put:  Is it OK to have the use/non-use of ICE as a fixed
attribute of each overlay?  This proposal is because there is no known
way to make mixed ICE/non-ICE use within one overlay to work.

No objection.

slide:  Direct Return Responses

David Bryan asks:  Do we have problems with compression of via lists, retransmission, and direct return?

Cullen:  I think it's been handled effectively.

Roni Even:  Intermediaries need to time out transaction state even if
they don't see responses.

---> Note for Cullen:  Make sure it is clear to implementers how to
     time out state in intermediate systems handling requests.

Question put:  Any objections to closing the issue?

No objection.

slide:  Overlay Algorithm

Question put:  Shall we add a flag as to whether sender wants the
finger table in the response?

No objection.

slide:  Transport Security

Proposal:  Propose to have pluggable security mechanism, with TLS
being the initially defined mechanism.  An overlay's configuration
will specify the security mechanism it uses.

Roni Even: Asks why, if IPSEC is the chosen security mechanism, it
would be necessary to specify it, since IPSEC operates at a lower
layer than the application.

Cullen:  Confirming the identity of the far end requires interaction
between the application and IPSEC.

---> Note for Cullen:  Need to revise text to make it clear that the term
     TLS is being used as a placeholder meaning "the configured security
     mechanism" instead of only TLS. The term TLS will not be changed in
     every location.

Chair:  That draft is close to being complete regarding security
configurability and people who have complaints should speak up.

No objections.

slide:  Peer-ID Length

Proposal:  Length of the peer ID is fixed per-overlay, with a maximum
of 160 bits.

No objections.

David Bryan:  Requests document be updated to clarify that "opaque"
objects self-describe their lengths.

---> Note for Cullen:  Promises to clarify in the draft that "opaque"
     objects self-describe their lengths.

slide:  Revision Plans

Chairs ask for volunteers to do reviews of document before WGLC.

----------------------------------------------------------------------

Haibin Song presents draft-ietf-p2psip-diagnostics-04:
P2PSIP Overlay Diagnostics

slide:  Changes since last version - 1

slide:  Changes since last version - 2

slide:  Changes since last version - 3

slide:  Changes since last version - 4

Chair:  Intention is to WGLC this document immediately after WGLC for
the Reload draft (draft-ietf-p2psip-base).

----------------------------------------------------------------------

Alexander Knauf presents draft-knauf-p2psip-disco-00:
A RELOAD Usage for Distributed Conference Control

slide: Outline

slide: Problem Statement for Conferences in P2PSIP Scenarios

slide: Objectives of Distributed Conference Control

slide: Distributing a focus with SIP

slide: Operations in a distributed conference

Brian Rosen: If each participant is attached to one (distributed)
focus peer, what happens if a focus peer fails?

Knauf:  Each participant attached to that focus peer hunts for another
focus peer through the usual discovery mechanism.

slide:  Definition of a Distributed Conferencing (DisCo) Kind

slide:  Creating a conference

Cullen Jennings:  A lot of original design was premised on permitting
continuing functioning when the credential server fails.  But this
design requires the credential server for ongoing operation.  I
encourage adjusting the design to not require the credential server
for operations.

slide:  Joining a Conference and publishing Focus-ability

Brian Rosen:  This mechanism has more uses than conference focuses.
Can the document be structured so as to separate the definition of the
general mechanism and the specific use for conference focuses.

Eric Rescorla: What if the key is stolen?

Gabriel Hege:  We haven't considered it.

Jin Peng:  If a focus peer fails, can another participant take over
doing signaling and media mixing?

Alexander Knauf:  Haven't worked out how to recover media mixing.

Roni Even:  Who establishes the media parameters of the conference?

Gabriel Hege:  The first peer to join establishes the media
types/codecs.