Minutes: XDI TC Telecon Friday 2012-03-02

[Drummond Reed <drummond.reed@xdi.org>]

Following are the minutes of the unofficial telecon of the XDI TC at:

Date:  Friday, 24 February 2012 USA

Time:  9:00AM - 10:00AM Pacific Time (17:00-18:00 UTC)

ATTENDING

Bill Barnhill

Joseph Boyle

Markus Sabadello

Drummond Reed

Giovanni Bartolomeo 

REGRETS

Mike Schwartz (at OpenID Connect interop at RSA)

THE IDEARPAD LINK FOR TODAY IS:

     http://xdi.idearpad.org/56

1) IDTRUST BUDGET PROPOSAL

Markus said that Scott McGrath is reviewing the proposal.

2) OPEN SOURCE PROJECT UPDATES

Drummond reported that Mike Schwartz was not able to attend today because he was showring the OpenXDI OpenID Connect server at the OpenID Connect Interop at RSA. The OpenID Connect server is a separate server in the OpenXDI suite. In a meeting earlier this week he demonstrated to Drummond that it was receiving the highest interop scores of all participants.

http://osis.idcommons.net/wiki/OC3:Cross_Solution_RP_x_OP_Results

Markus said he's still experimenting with p2p architectures. 

Drummond said that he attended a review of OpenXDI with Mike and Phil Windley earlier in the week.

Bill said that EzXDI has basic XDI operations are working using JWE (JSON Web Encryption) and JWS (JSON Web Signature). JWE and JWS were developed in the context of OAuth and are being used with OpenID Connect. Bill said that he's using Markus' XDI Squared code as his initial interoperability code.

# DRUMMOND to add OpenID Connect question session with Mike to the agenda for next week.

3) PROTOCOL DISCUSSION

We discussed XDI protocol implemention options. Bill said that he's developed a protocol for XDI over WebSockets, which  is supported by all the major browsers. This has some advantages in terms of supporting push notifications, and also in efficiency vs. HTTP Post. HTTP Post would become a fallback method of the WebSockets method.

Giovanni asked whether this meant that we were moving away from the REST architectural model. Bill answered that WebSockets for XDI messages is essentially bi-directional REST, where both the clients and the servers can exchange XDI graph state. 

Giovanni summarized that Bill's proposal is essentially using HTTP to setup the channel for XDI messaging. Bill agreed, and said this same pattern could be applied to JMS, XMPP, and AMQP (the queuing protocol using by RabbitMQ).

Giovanni suggested that we will still need to map the XDI protocol to "pure HTTP REST" for interop with standard HTTP connections. 

Bill said that if we want to do this, we'll need to "package" sets of XDI statements that would ordinarily be independent as individual resources. One way of doing would be to orient it around XDI messages being the resource. While that might work, REST purists may insist that the resources be subgraphs. This would require each HTTP message to be an atomic operation on the graph, which would be much less efficient, but would meet the pure REST requirements.

Drummond proposed that the TC concentrate first on Bill's proposal to do a first binding to HTTP/Websockets that falls back to HTTP, and then a second binding to pure HTTP REST. In the end, both protocols will transmit all state changes to the XDI graph, and both philosophical camps will satisfied.

# DRUMMOND or BILL to reflect this in the XDI Specifications wiki page.

4) SPEC REPOSITORY REPORT

Bill hasn't made any movement forward on this, but expects to have progress in the next two weeks.

5) LITERAL CONTEXTS & BINARY DATA

We had a short time to discuss this thread about how literal contexts solve datatyping issues:

  http://lists.oasis-open.org/archives/xdi/201203/msg00003.html

In particular we talked about binary encoding. Drummond explained that the that literal contexts provide all the metadata about the type of a literal, including binary data.

We discussed how binary data shoud be serialized for transmission in XDI messages. The conclusion was that we should define a single encoding for binary data in our JSON serialization spec, and that encoding should be used for all binary data.

The topic led us to talk about the XDI type dictionary, and the need to establish canonical XRIs for each datatype that will be defined.

Bill said that in creating the XDI type dictionary, we should give MIME types first precedence. Drummond suggested that we go in historical order, i.e., MIME types first, XSD second, and JSON third.

6) NEXT CALL

The next call is next week at the regular time. Drummond, Joseph, and Mike should all be able to do the call together on the first day of SXSW.