Behavior of Span.End is unclear.

[Oberon00]

The behavior of Span.End could use some clarifications. For example, I think I heard from @tsloughter that in Erlang, ending a Span inactivates it on a Context there, which seems problematic, especially after Context is the only way to specify a parent span. Other points that are unclear are when a span needs to be ended, and what the relation to IsRecording is.

I created PR #1011 to clarify these points.

[tsloughter]

So, yes, we don't have users use Span.End directly in Erlang/Elixir, similar to why Java uses Scope instead of requiring a user to explicitly call End and then manually change the active span to the previous active Scope's span.

Instead users use ot_tracer:end_span(Tracer) which ends the active Span in the current context.

This allows for Tracer to be responsible for Span Processors, if it wasn't then we'd have to copy Span Processors from the Tracer into each started Span.

The Tracer's end_span calls the Span modules end_span, runs the OnEnd processors and then makes the Span that was previously active now active and the current active span no longer active.

So it isn't a function of Erlang's Span.End that inactivates the Span in the Context but a function "above" that, currently in the ot_tracer module.

[andrewhsu]

from the issue triage mtg today with TC, allowing changes related to this issue for GA if editorial changes only

[tsloughter]

I'd be interested in seeing how other languages allow users to end a span and return to the previously active span. I don't think I've seen any that requires the user do:

PreviousSpan = Tracer.get_active_span()
Span = Tracer.start_active_span()
Span.End()
Tracer.set_active_span(PreviousSpan)

[tsloughter]

I guess what I'm suggesting is adding an End in "Tracer Context Utilities" that ends the active span and updates the context. But maybe only Erlang "needs" this?