Glossary term Event needs more explanation

[chalin]

Context https://cloud-native.slack.com/archives/C01N7PP1THC/p1737719337217569:

... can someone help me to find or explain the difference between events and logs? https://opentelemetry.io/docs/specs/semconv/general/

This should be clarified in the Event glossary entry IMHO.

In particular, we might also want to link to:

• https://opentelemetry.io/docs/specs/otel/logs/data-model/#events

WDYT?

/cc @open-telemetry/spec-sponsors

Related:

• Editorial changes to Concepts > Signals #6067

[Kludex]

Can we clarify the use of span events and log records for exceptions? Currently, at least in Python, the opentelemetry-python sdk uses span events to record exceptions, but it seems the spec says that we should use a log record.

[adriangb]

I also would like a basic clarification of the relationship between span events and the "events" (the ones that are seemingly being replaced by logs).

[adriangb]

I also would like a basic clarification of the relationship between span events and the "events" (the ones that are seemingly being replaced by logs).

https://github.com/open-telemetry/opentelemetry-specification/blob/main/oteps/0265-event-vision.md#relationship-to-span-events

Events are intended to replace Span Events in the long term. Span Events will be deprecated to signal that users should prefer Events.

Interoperability between Events and Span Events will be defined in the short term.

[pellared]

Here is how we described it in OTel Go:

A log record with non-empty event name is interpreted as an event record.

You can think of "event.name" like a "log record type".

From the OTel world, I think that we could have span.started events (log records with event name span.started) with the information about the started span. This could e.g. help users that experience long-living spans or people who do not have a tracing backend (like Jaeger) and have to rely on logs.

[adriangb]

From the OTel world, I think that we could have span.started events (log records with event name span.started) with the information about the started span. This could e.g. help users that experience long-living spans or people who do not have a tracing backend (like Jaeger) and have to rely on logs.

We've already implemented this, it works ~ generally well, but there are gotchas.

[pellared]

CC @open-telemetry/semconv-event-approvers

[trask]

Can we clarify the use of span events and log records for exceptions? Currently, at least in Python, the opentelemetry-python sdk uses span events to record exceptions, but it seems the spec says that we should use a log record.

the spec doesn't say this yet, though we're going in that direction: open-telemetry/opentelemetry-specification#4333