Develop strategy for deprecating / removing experimental semantic conventions
jack-berg opened this issue · 3 comments
We currently maintain many outdated attributes with @Deprecated annotations and previously agreed that this is unsustainable.
For stable semantic conventions, we should obviously retain attributes indefinitely, even if they undergo allowed renames via schema transformation rules.
However, for experimental conventions, we should develop some procedure for end-of-lifing attributes as the conventions churn.
Once we develop this strategy, we should consider removing all current attributes marked as deprecated.
I think as long as the prior @Deprecated version indicates what the new thing is called or what the user should use instead, then it should be safe to remove these.
Can't the "procedure" to remove these just be a simple as removing them after a release cycle has passed?
EDIT: Ok, I think I understand now -- the process suggested here is intended to help automate initiation of the deprecation itself.
Created generalized issue in the semconv repo - open-telemetry/semantic-conventions#550
Python has the same problem and it'd be great to have language-agnostic solution
This is resolved. See #50 (comment) for a full summary of the strategy.