Consistent use of diagnostics_channel events to demarcate the boundaries of a request #4
Description
I've been working on building up diagnostics_channel and TracingChannel as a high-performance data source for tracing products to consume safely rather than relying on risky monkey patching. I would like to extend that effort to defining consistent events for serverless function providers to publish around the lifecycle of a request to form a trace boundary.
For the tracing use case a clear start and end event will be needed to scope the activity of that request. Additionally some error tracking event would be valuable. This could be a single event or it may be worth splitting out certain commit failure cases such as timeouts.
Is this something others feel makes sense as a part of this functions API initiative, or should standardization of this debugging information be considered separately? What are others thoughts on what this should look like?