Minutes IETF99: tokbind
minutes-99-tokbind-00

Token Bindings IETF 99
===

0RTT and 1RTT
---

- Nick presents 0RTT and 1RTT discussion
  * Lucy & Martin expresses support for multiple drafts
- Nick presents changes from -01
  * Martin asks question about DHE KEX - may need to open issue to ensure
  uniqueness * Nick clarifies the threat-model in play for DHE KEX
- Nick presents overview of 0-RTT TB initial handshake
- Nick presents exporters requirements for 0-RTT
  * Martin: protocols should specify a profile both for 0RTT and 0RTT-TB
  * Martin: include section that contains the profile for 0RTT-TB for HTTP
  * Martin: include advice for other protocols
- Nick presents options for switching exporters
  * Consensus around using normal exporter
  * Discussion at mic on the value of the client certificate analogy

Proxies and Terminators
---

- Brian intro to TTRP draft based on feedback from IETF98
  * MikeJones: have usecases that require > 2 token bindings, pls specify
  syntax that allows multiple TBs * StefanSantesson: have you considered AJP? *
  Brian: no * EKR: sanitization seems sketchy * Brian: aware of feedback *
  Brian and EKR discussion the header security / sanitization issue * Dirk:
  based on google experience support the current proposal * ERK: why not
  establish a shared key and MAC the header * Brian: security considerations
  support that notion * EKR: risk of misconfiguration is signifficant - if the
  orig headers are passed thourough => major issue * Brian: worried about key
  mgmt for one specific application - need broader applicability * JohnBradley:
  what wg would be appropriate for a general solution? * MikeJones: ... *
  JoeSaloway: sanitization is a problem - needs a solution * WilliamDennis: not
  fully understand the problem.. more complexity adds risk that TB doesn't get
  deployed * (?)Google: passing EKM is better than new shared key mech *
  RichSaltz: assumption of trust in TTRP is not valid * MartinT: its simple to
  implement MAC using TLS exporter on the inside leg * EKR clarifies as AD that
  this is a WG decision * Discussion around the assumptions - is TLS used * LJ:
  how can we produce a document based on MartinThomsons ideas? * Brian: TTRP is
  important for getting deployment of TB * various folks involved in a
  discussion on where and when to create a MAC spec * WilliamDennis: complexity
  should be at the operator/data center level * LucyLynch: appreciate draft for
  exposing important issues: sanitization etc

  * Brian contiues preso
  * MartinT: don't provide information to backend that the backend can't use
  * MartinT: MikeJones request for > 2 TB IDs is fine
  * MartinT: make these Sec-* headers
  * StefanSantesson asks clarification on the sanitization issue
  * Vinnod: why not provide both TB ID and EKM for more complex applications?
  * Nick & Brian & MikeJ discuss requirements for > 2 TBs

  * Brian shows example of running code

- Open Mic:
  * StefanSantesson: TB for SAML?
  * MikeJones: should use confirmation method (borrowed from SAML) based on the
  work in OpenIDC * MikeJones: no opinion as to where it gets done * JeffH:
  agree with Mike - * There seems to be support...

  * Lucy: Nick - please include a problem statement