Opentelemetry-specification: Improve error handling

[tedsuo]

What are you trying to achieve?
Improved error handling, based on OTEP 136.

• Reduce the number of Status Codes.
• Add Status Source.
• Indicate which semantic conventions generate errors.

Additional context.
OTEP PR: https://github.com/open-telemetry/oteps/pull/136

This issue supersedes https://github.com/open-telemetry/opentelemetry-specification/issues/706.

[arminru]

Classified as p1 since it supersedes the p1 #706 and is quite pervasive.

Indicate which semantic conventions generate errors.

Should we split this into a separate (p3 or after-ga) issue?

[tedsuo]

@arminru I will include the HTTP status codes as part of this issue. I agree a separate issue for the rest of the semantics.

[carlosalberto]

@tedsuo @arminru my impression is that we should close this issue, as the core has been implemented/updated - the remaining semantic conventions to be updated can be labeled as "good do have for GA", as they technically won't break the API ;)

EDIT: I just saw that indeed we plan to create a new issue for the follow up ;)

I agree a separate issue for the rest of the semantics.

[carlosalberto]

Closed via #966

[arminru]

Follow-up for the semantic conventions: #1003

[arminru]

@open-telemetry/technical-committee @tedsuo

I just noticed this issue was closed without adding Status Source (instrumentation | user), which was part of the OTEP and also the issue description ("Add Status Source.") to the spec in #966.

Was this missed or intentionally dropped from GA?

[carlosalberto]

@arminru IIRC StatusSource was removed based on public feedback. @tedsuo May remember the details ;)

(And to be clear: the OTEPs can always still change when hitting the Specification. A sampling and a context-prop OTEP come to mind).

[Oberon00]

I think the comment that successfully argued for the removal is this: https://github.com/open-telemetry/opentelemetry-specification/pull/966#discussion_r490586553

As pointed also in the otep. I am worried that we add things right before GA that we will regret and have to maintain for years.

I strongly suggest to not add the source yet, adding it later is backwards compatible.

[arminru]

I added #1092 (labeled as after-ga) to keep track of this part.