You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Today we require all attribute definitions to happen in the registry, i.e. :
attribute definitions should be in /model/registry
the group id must start with registry.
We should consider having a more formal way to mark a group as one defining something.
E.g.
groups:
- id: foo.bartype: attribute_registry <-------- this is new
or
groups:
- id: new-guidtype: metric_registry <-------- this is new
This would simplify filtering/massaging semconv in code-generation, we'd avoid relying on specific group naming.
This would move post-factum policies enforcing attribute definitions into schema validation (id can only appear on *_registry groups and ref cannot appear there).
The text was updated successfully, but these errors were encountered:
It's a good question if we can reuse existing types (attribute_group, metric, etc) and I believe the answer is no:
we use attribute_group a lot to define a common group of ref attributes that are reused between multiple spans or metrics. This helps to avoid writing repetitive notes and similar for different http/messaging/etc metrics. E.g.
Today we require all attribute definitions to happen in the registry, i.e. :
/model/registry
registry.
We should consider having a more formal way to mark a group as one defining something.
E.g.
or
This would simplify filtering/massaging semconv in code-generation, we'd avoid relying on specific group naming.
This would move post-factum policies enforcing attribute definitions into schema validation (
id
can only appear on*_registry
groups andref
cannot appear there).The text was updated successfully, but these errors were encountered: