search

open-telemetry / opentelemetry-js

OpenTelemetry JavaScript Client
https://opentelemetry.io
Apache License 2.0
2.62k stars 759 forks source link

@opentelemetry/semantic-conventions out of sync. E.g. 1.25.1 `http.user_agent` does not follow semantic conventions? #4843

Closed AkselAllas closed 1 month ago

AkselAllas commented 1 month ago

What happened?

In semantic-conventions 1.19.0 there is this line:

Rename http.user_agent to user_agent.original

But I have span from instrumentation-http 0.52.1, which uses 1.25.1 of @opentelemetry/semantic_conventions with http.user_agent (defined here) which is not what semantic convention defines?

Am I misunderstanding something on how these versions are supposed to work?

Span: image

OpenTelemetry Setup Code

No response

package.json

No response

Relevant log output

No response

AkselAllas commented 1 month ago

Seems semantic-conventions is in violation of this line from docs:

The YAML files MUST be used as the source of truth for generation.

pichlermarc commented 1 month ago

Hi, thanks for reaching out.

Seems semantic-conventions is in violation of this line from docs:

The YAML files MUST be used as the source of truth for generation.

The semantic conventions are out of date, but they do follow the YAML - just a very, very old verison of it. @dyladan is working on updating them. See #4572, #4690, it's a rather complex situation as we released 1.x which exported experimental attributes as stable, so we can't just drop these.

This issue is a possible duplicate of #4572

AkselAllas commented 1 month ago

I just wanted to say that as an outside viewer, I expected 1.25.1 release of semantic-conventions to actually follow the YAML of 1.25.1 version.

JamieDanielson commented 1 month ago

I just wanted to say that as an outside viewer, I expected 1.25.1 release of semantic-conventions to actually follow the YAML of 1.25.1 version.

That is a fair point. In fact since we are working on semantic conventions now and the version numbers are close to matching up, I think we are going to separate the publishing of the semantic conventions package so we have more control over the version number. This will hopefully make it more clear going forward.

In the meantime I will close this issue as a duplicate being tracked elsewhere for updating semantic conventions.