Hi folks,

<speaking as draft author and not chair> 

If you recall from the presentation at IETF 119, the authors have been working 
hard to ensure CDDL types are used effectively. Part of that work is in 
supporting broad extensibility with structured rigidity. 

The main schema defines that each event has a generic "data" field that is 
intended to be specified by the use case of the event. For instance, the QUIC 
event schema extends this field to accomodate various QUIC related events. 
We've had an issue open for a while to tighten this extensibility up [1]. We 
have a PR for that [2], which has general acceptance amongst the authors. 
However, an offshoot of this design change is that it drops the "trigger" field 
that had nominally been a part of every event. 

My own qlog library implementation had always ignored this generic trigger 
field, so personally I see dropping it from the specification as a positive. 
From an application perspective, it can be difficult to propogate a reason to 
the right place where the qlog would be generated. There can also be privacy 
matters to contend.

Removing the top-level trigger now doesn't mean it can't be added by later 
extensions, or as an optional private/custom field. For example, the QUIC 
connection_closed event [3] has it's own specific trigger field with a set of 
pre-defined values.

Before we go ahead with merging the PR [2], we wanted to surface this change a 
bit in case other implementations had built a reliance on it. Please respond on 
the issue, or to this thread, if you see some issue.

Cheers
Lucas
On behalf of the qlog document authors


[1] - https://github.com/quicwg/qlog/issues/192
[2] - https://github.com/quicwg/qlog/pull/401/files
[3] - https://quicwg.org/qlog/draft-ietf-quic-qlog-quic-events.html#section-4.3


Reply via email to