Require that names and namespaces are one global space across all semantic convention areas #832
Conversation
tigrannajaryan
added
the
area:semantic-conventions
|
Please rebase @tigrannajaryan |
specification/common/common.md
Outdated
| namespace. Because of this rule be careful when choosing names: every existing | ||
| name prohibits existence of an equally named namespace in the future, and vice | ||
| versa: any existing namespace prohibits existence of an equally named | ||
| attribute or label key in the future. |
There was a problem hiding this comment.
I agree with the overall uniqueness requirement, but the motivation for this paragraph is not clear to me. Why would it be bad to have a name service.instance? Yes, it clashes with the namespace of service.instance.id, but so what?
There was a problem hiding this comment.
Often the namespaces represent "things". For example all attributes in k8s.pod.* namespace are about Kubernetes Pod. If we allow an attribute name that coincides with a namespace it appears that the value of the attribute describes the "thing". However, in this case why is it not just another attribute under that namespace? What would be a reasonable value for k8s.pod attribute that does not have a better name in the k8s.pod.* namespace?
I think if attributes like that existed it would be a source of confusion. Does service.instance attribute describe something about service (as a "thing") or something about service.instance (which is a different "thing")?
The requirement I added prevents this confusion. When attributes are used for "things" it is very clear which "thing" they are about, it is the "thing" denoted by their namespace (whereas if we allowed what you suggest it could be also the "thing" denoted by the parent namespace).
…antic convention areas We have semantic conventions for Resources, Spans and Metrics (in the future also Logs are expected). It was not clear if the attribute names across all convention areas should be globally unique. This commit asserts that conventions are one space, they are not independent spaces with their own namespaces each. We prohibit using the same Span or Resource attribute name or metric label name but give them slightly different meanings or value sets. Resolves: open-telemetry#815
tigrannajaryan
force-pushed
the
feature/tigran/unique-semantic-conv-area
branch
from
da6923b
to
e05488a
Compare
…antic convention areas (open-telemetry#832) * Require that names and namespaces are one global space across all semantic convention areas We have semantic conventions for Resources, Spans and Metrics (in the future also Logs are expected). It was not clear if the attribute names across all convention areas should be globally unique. This commit asserts that conventions are one space, they are not independent spaces with their own namespaces each. We prohibit using the same Span or Resource attribute name or metric label name but give them slightly different meanings or value sets. Resolves: open-telemetry#815 * Address PR comments Co-authored-by: Bogdan Drutu <bogdandrutu@gmail.com>
We have semantic conventions for Resources, Spans and Metrics (in the future also Logs are expected).
It was not clear if the attribute names across all convention areas should be globally unique.
This commit asserts that conventions are one space, they are not independent spaces with
their own namespaces each.
We prohibit using the same Span or Resource attribute name or metric label name but give them
slightly different meanings or value sets.
Resolves: #815
This PR is stacked on top of #821
Ignore the commits from #821
Only review commit f4f9555