Christian Neumüller

I feel like retry & redirect should be counted with the same counter (they are both "re-sends", see https://www.rfc-editor.org/rfc/rfc9110#name-redirection-3xx). One could add something like a "resend_reason" that, apart from distinguishing...

> Question: should the unreleased changes considered to be official part of the spec or they only become part of the spec once the release is made? I think the...

* Body can be a semantic convention for attributes. * Severity can be a semantic convention for attributes * ShortName could be renamed to Name. * Why are nested attributes...

> If we do re-use the traceID from untrusted sources we have bigger problems than the sampler. I think that topic should be addressed independently and separately. I think you...

I oppose this change. Why? Because I think there should be a single resource attribute that tells you from which kind of process/runtime any telemetry is emitted from. Of course,...

I see telemetry.sdk.language has been repurposed to also distinguish nodejs and webjs even though both use javascript. I did not know that. Still, that doesn't really change my argument. If...

A concern in the spec SIG meeting that came up was, for which strings this should apply: * Attribute and label keys? * Span and event names? * Metric label...

People (incl myself 😉) are putting 😄-reactions under @tsloughter's comment, but these are all real problems. Unicode is complex. > require that the truncation happen on grapheme cluster I think...

And there might be cases where you don't want to normalize, e.g. when you implement metrics for some Unicode normalization microservice (just pulling this example out of thin air here)...

So that `流量` becomes `__`? Does not sound very helpful. We could use urlencoding to at least avoid introducing collisions. But I guess there is only so much we can...