Closed MrAlias closed 1 year ago
I'm going to pick this up!
Note on the spec: It is a bit odd, IMO, for advice to be identifying for instruments (see the instrument spec: https://github.com/open-telemetry/opentelemetry-specification/blob/4ac43da9d746fea42523187685f341321a1d6ee1/specification/metrics/api.md#instrument). Does that mean the same instrument could be created, but with different buckets?
Does that mean the same instrument could be created, but with different buckets?
Hmm, yeah, that seems like an incorrect design.
Alt prototype: https://github.com/open-telemetry/opentelemetry-go/pull/4340
Both prototypes have been reviewed to satisfy the issue.
Other feedback for the spec: Advice seemed unneccessary. I can't see a world in which I would ever want to set histogram bounds on a non-histogram instrument. If someone is configuring the aggregation in a view (from a non-histogram instrument), they need to set the buckets anyways, so the advice can't be used.
Adding the explicit bucket histogram advice API is tracked here: https://github.com/open-telemetry/opentelemetry-go/issues/4003
In order to make a stable release of the metric SDK, it needs to be ensured that that API can be added in a backwards compatible manner. This issue tracks that verification.
Resolving this may result in a complete solution being submitted as a PR, but it could also be resolved with a clear description of how to add a compliant implementation in a backwards compatible manner.