W3C

– DRAFT –
Automotive Working Group Teleconference

19 October 2021

Attendees

Present
Adnan, Ashish, Erik, Gunnar, Isaac, MagnusG, Peter, Ulf
Regrets
-
Chair
Peter, Ted
Scribe
Ted

Meeting minutes

MQTT

https://github.com/w3c/automotive/issues/369

Ulf: proposal I made earlier wasn't regular MQTT interface but an alternative that exposes the functionality of CORE spec over MQTT similar to how you can interact over WSS and HTTPS
… I've implemented this as well, back in February
… there was some pushback at having a protocol on top of MQTT and wanted to get some MQTT experts to provide more input
… I find value in having VISS over MQTT instead of using their more traditional route with eg broker

Gunnar: I think it is perfectly fine emphasizing same is being done with WS and HTTP

Isaac: this might affect the security model of MQTT
… with that protocol, the different points would be talking through the broker
… we use long term tokens and could allow them to be used across protocols

Ulf: good point, wonder if that would be better as a feature for next release
… or as Isaac describes, we can do our own end to end encryption using the long term access control mechanism

Gunnar: perhaps later version indeed and based on interest of participants. with WS and HTTP we require use of TLS

Erik: would it be worth identifying use cases where we are fine without encryption?

Ulf: agree with Isaac the weak point in MQTT is the broker but one can presume it is well secured in vehicle implementations

Isaac: we need to emphasize importance of being sure where MQTT broker is running, maybe give a few scenarios where it would be acceptible

https://www.w3.org/community/autowebplatform/wiki/Best_Practices

Ted: suggest longer portions, eg example architecture scenarios where it is safer would belong on BP

Ulf: so are people alright with me proceeding with this proposal?

Gunnar: including making clear the risk/concern?
… how broker could expose data for instance

Ulf: alright, I'll try to write this

Isaac: willing to help

Issues and PR

https://github.com/w3c/automotive/issues/382

Ulf: currently in the model we have role based system and a purpose model requiring a policy document
… the access token are not explicitly shown but purpose, scope is provided and policy document used to map to specific signals allowed
… there was some discussion about having signals in the access token itself
… a simpler auth flow the client has received a token out of bounds of the spec
… this Signal Access Control (SAC) token would include explicit signals and remove the role based part along with policy document

Isaac: I meant to contribute to the issue based on previous discussion but forgot

Ulf: Erik, I encourage you to take a look at this proposal since we did so responding to desire to make an alternative path per Sebastian's request

Isaac: we could perhaps keep the access token server

Ulf: if you could write that please

Isaac: noted

Ted: ideal would be transparent to client application which auth path is provided

Isaac: there may be a couple alternatives to discuss

https://github.com/w3c/automotive/issues/422

Ulf: question is whether we can have multiple purposes in the same token

https://www.w3.org/2021/10/12-auto-minutes.html#t02

Isaac: it is difficult to handle multiple tokens within a client

Ulf: the access grant token contains the authorized roles. with that single token it is possible the handle multiple purposes for that specific role
… one scope claim is possible

Isaac: if requesting multiple signals across purposes would you be able to send multiple tokens?

Ulf: no

Ted: two ideas, customizable new roles for specific purposes

[ick]
… or SAC we were just talking about for alternate auth path of previous issue

Ulf: interesting
… SAC could handle tailored access

Isaac: it could make sense

Ulf: I'll provide a comment to this issue

Minutes manually created (not a transcript), formatted by scribe.perl version 136 (Thu May 27 13:50:24 2021 UTC).

Diagnostics

Succeeded: s/path/auth path/

Maybe present: Ted