Thank you Dominik for your review of the draft. We mentioned in the virtual
interim that we approached to OASIS MQTT TC for reviewers; Dominik  kindly
agreed to review our draft.

Indeed, the PINGREQs were not included for checking token expiry as they do
not signal a token (like CONNECT) or topic permissions (like
PUBLISH/SUBSCRIBE). But, your suggestion would be a very useful
optimization for early detection of token expiry. As Anthony said, we will
add it to the draft (possibly as a SHOULD).

Very glad to hear that your ongoing project is in agreement with the draft.
It would be very useful to get input from your deployment experience.

Sincerely,
--Cigdem

On Wed, Nov 8, 2017 at 1:34 PM, Anthony Kirby <[email protected]>
wrote:

> Hi Dominik,
> many thanks for the review! It's appreciated.
>
> We didn't initially consider the keepalive / PINGREQ, because we'd seen it
> as (effectively) a lower layer issue.  But I understand value of using it
> as an opportunity to check for token expiry:  Even if it happens
> frequently, each check will be a low cost for the broker.  Thank you for
> the suggestion, I'll add it to the draft.
>
> Thinking aloud, it might not be straightforward to implement this when
> extending an existing broker (for example the mosquitto auth plugin) so I
> expect it would be a SHOULD rather than a MUST.
>
> Anthony
>
>
> ________________________________________
> From: Ace <[email protected]> on behalf of Dominik Obermaier <
> [email protected]>
> Sent: 07 November 2017 21:29
> To: [email protected]
> Subject: [Ace] MQTT-TLS profile of ACE
>
> All,
>
> I had the opportunity to review the MQTT-TLS profile for ACE [1] and
> I’d like to share my thoughts with this list.
>
>
> The document at hand proposed that token expiry checks take place on
> PUBLISH, SUBSCRIBE or CONNECT actions. I’d like to note that it might
> be worthwhile to have the token expiry checks for PINGREQ packets in
> order to make sure that “sleeping” MQTT clients which do not send
> PUBLISH or SUBSCRIBE packets can get disconnected at the next possible
> client interaction with the broker.
>
> On a sidenote: In a recent MQTT project with millions of constrained and
> untrusted devices (connected via unreliable communication channels), an
> almost identical approach as proposed in the draft was implemented. The
> authentication and authorization was implemented exactly as described in
> this document with the use of the Introspection API and “offline
> validation” of JWTs. So I can confirm that the approach proposed is
> actually usable at scale and works very well with some existing MQTT
> brokers.
>
> All the best,
> Dominik
>
> [1] https://www.ietf.org/id/draft-sengul-ace-mqtt-tls-profile-01.txt
>
> _______________________________________________
> Ace mailing list
> [email protected]
> https://www.ietf.org/mailman/listinfo/ace
>
> _______________________________________________
> Ace mailing list
> [email protected]
> https://www.ietf.org/mailman/listinfo/ace
>
_______________________________________________
Ace mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/ace

Reply via email to