draft-ietf-behave-nat-behavior-discovery-02

1
draft-ietf-behave-nat-behavior-discovery-02

• Derek MacDonald
• Bruce Lowekamp

2
Changes from IETF-69

• Changed intended status to experimental, based on
  list discussion.
• Per discussion at 69 and changes to 3489bis
• Removed mandatory shared secret, either use
  rate-limiting or credentials
• 3489bis compatibility addressed by making
  attributes optional, renaming attributes that now
  have different numbers.

3
Applicability

• Clarify not intended to duplicate original 3489
• Two uses described
• Diagnostics
• Intended for administrator/programmers
• Not considered experimental
• Application tuning
• Example of P2P role selection
• Describes experimental aspects (on next slide)

4
Experimental Use P2P Overlay

• Proposed as specific example of experimental test
  case.
• Perform behavior-discovery using other peers
  collectively as stun server
• Determine whether able to be a peer or client (or
  whether to provide TURN service) based on results
  and contact with other peers
• Used for initial selection of parameters.
  Choices will likely vary with further experience.
• Key question does behavior discovery provide
  less likelihood of costly mistakes than other
  approaches?

5
Criteria for Success

• An experiment would be considered a success if
• An implementation determines its initial choice
  of behavior based on behavior-discovery knowledge
• Must operate either in a deployment with STUN
  servers or duplicating a STUN servers behavior
  with multiple peers
• Must present experimental evidence that the use
  of behavior-discovery results in improved
  performance and network/application stability
• A protocol specification must define how
  behavior-discovery is applied