We should stick to our "division" of "who" and "how" uses this API and fully document it.
Only once fully documented and proven in one or two integrations can we slowly claim to stabilize the API.
Complete documentation has a magical power of uncovering edge cases etc,
and of course matters for adoption.
Our "personas" are:
end users
"pass around context" instrumented libs do the right thing
you can get a tracer and start spans yourself
library and framework authors / "system authors"
this is where and how to pass context
this is how you can start spans on behalf of users
We should stick to our "division" of "who" and "how" uses this API and fully document it. Only once fully documented and proven in one or two integrations can we slowly claim to stabilize the API.
Complete documentation has a magical power of uncovering edge cases etc, and of course matters for adoption.
Our "personas" are: