ABFAB Notes IETF93
===========

* UI Considerations
  + 3-minute summary
  + Document has been revved based on feedback
  + More comments forthcoming
  + Anticipate getting it finished this week
    - Leif hopes for WGLC by this Friday
    - And writeup stage before September
* AAA-SAML
  + Main issue: mapping between SAML names and AAA names
    - Proposal went out yesterday
  + Other relevant issues
    - Include a nomenclature table (done)
    - Change the name of "SAML Message Attribute" to "SAML Protocol"
    - Q: Are domain-only NAI representations allowed in the Network Access Identifier Name Identifier Format?
      - A1: NAI allows that
      - A2: If we don't allow that, we'd need another convention for anonymous naming, so might as well use the one that exists
    - Q: Should section 6.1 Confirmation Method identifiers also refer to the ones in section 8.1?
      > A: Jim got an answer from Scott, then forgot it
  + Proposal for wrapping SAML in RADIUS data
  + Proposal to name an AAA identity using a URI
    - Sam didn't like that.
    - Go through an example RP metadata
  + Sam at mike: Thanks to Alejandro for moving this forward
    - Registering a RADIUS URI? Might be difficult.
    - Want to discuss the metadata structure - if the registering were easy, would this be what we want?
      > Leif: This looks pretty close to what he would like
      > Leif: Do we need a role descriptor for this?  Use the one that already exists, but does it need a different binding?
    - Trying to register will be a lot of work
      > Especially for a URI scheme without hostname or port or such
      > We'll have to talk to APPS and RADExt more than we want to
      > Q: Could we add extra attributes?
        o Want to specify a placeholder URI for now, and replace it in the future if we need.
        o The endpoint type has a required binding location; response location is optional; local namespace attributes can be added
        o Propose to define a URN that indicates that location is context-dependant
        o Define a location that is "NO"
        o Attributes for NAI, NSID, GSS Name
        o This gets us out of defining a new URI
        o This also allows the use of a location URI, if someone comes up with it eventually.
        o Maybe put extensions on the RoleDescriptor types.
      > If RoleDescriptor doesn't need an endpoint, why don't we just get rid of the URN and stick it in the RoleDescriptor?
        o Leif: Yeah, they don't need endpoints
        o Alejandro thinks that he knows how to move forward on that
        o Sam is in favor
        o Jim Shaad: So, basically we're splitting the name into other pieces?
        o Leif: We're much more worried about binding the name to a key
        o Leif: We're going to have to do cross-review with SSTC / OASIS
      > Who will take this to the SSTC / OASIS?
* Extra scheduled work
  + Stefan Paetow's presentation
    - SSH as a use case
    - Use ABFAB to do domain transitions
    - Log in to one system, then use that system to access further systems
    - Like SSH agent forwarding or XXXX from OAUTH
    - Cheznet did a proof of concept of getting an EAP-TTLS token
    - Interesting question: Can we get enough people to do the work?
      > Need more than JISC and Painless Security people to get enough review
      > ... and implementation!
* Where to go from here
  + Credential delegation
  + Ephemeral keying
  + Do we recharter with a couple of extension proposals?
    > Does Stephen want to get rid of us desperately?
    > Well, not desperately...
    > Finish the current set of documents first
    > Want to be finished by Yokohama
    > Will people be at Yokohama?  Or Buenos Aires?
    > Maybe have an interim or virtual interim?
    > Another option is to close the WG and have AD-sponsored documents for the remainder
  + Consensus: Finish what we have, then have an interim meeting
* Open mike:
  + GSSWeb
    > A chrome and firefox extension coupled with an apache module to do
GSS-EAP authentication
  + Launch of the academic service in the UK
    > Corresponding GEANT project to do the same, turn it into a global service