host.id and device.id are confusingly similar to each other

[Oberon00]

This is a sub-issue or closely related to #1765 "host attribute key should be interchangeable with device".

The host.id and device.id attribute seem to describe the same concept. Both are optional string attributes in their respective resources.

host.id: Unique host ID. For Cloud, this must be the instance_id assigned by the cloud provider. Example: opentelemetry-test

vs

device.id: A unique identifier representing the device [1] Example: 2ab2916d-a51f-4ac8-80ee-45ac31a28092

Footnote [1] of device.id does add a bit of distinction since it only allows certain values and then only defines them for iOS and Android, so it seems to be applicable only to these. Nevertheless, this could be replaced by requiring the use of these particular IDs on these particular platforms for host.id.

Otherwise a clarification should be added when to prefer which attribute and when to use both.

[bryce-b]

A lot of the attributes in the host semantic convention can be applied to a mobile device. There is already a semconv for devices and they have a lot of overlap.
There are potentially 3 things that could be done to flesh this out:

• Merge these two specs and allow them to be more interchangeable.
• Extend the device spec with missing and applicable attributes that already exist in the host spec, but in practice you wont see them side-by-side in the payload
• treat device and host independently. some meanings may overlap, any maybe used by implementers interchangeably, but the spec doesn't explicitly define that interchangeability.

[vvydier]

Can the device type not be captured in host.type for mobile? There are other mobile specific attributes (like the network type) that need to be added.

[ocelotl]

I asked Claude to review this issue and Claude suggests it be closed with this explanation:

Purely a resource semantic-convention clarification (labels area:semantic-conventions, spec:resource) about host.id vs device.id. All resource semantic conventions (host, device) migrated to open-telemetry/semantic-conventions, so this belongs there, not in the spec repo. Its only cross-reference, #1765, was closed COMPLETED and merely redirected discussion here; no resolution occurred and the discussion fizzled in 2021.

To guard against a mistaken suggestion, Claude then independently re-verified this conclusion:

Resource semconv (host.md/device.md) migrated to open-telemetry/semantic-conventions where these attrs now live; parent #1765 closed COMPLETED and this 2021 thread fizzled with no decision, so it belongs there not the spec repo.

Claude evaluated this issue with medium confidence that it can be closed.

Claude has classified this issue as belonging in another repository.

The intention of this comment is to make it easier for a human being to review this issue before making a decision.

[svrnm]

Closing this, as this is now a semantic conventions issue, if this is incorrect, please open a new issue on the semantic conventions repository.