Change w3c correlation context to custom header

[dyladan]

After discussions with the W3C Distributed Tracing Working Group, it has been determined that the Correlation-Context header is not ready for implementation and is being actively changed. As recommended by the working group, we should implement our own header until the specification is an official recommendation.

/cc @tedsuo @mwear

[bogdandrutu]

I am fine with this. QQ: should this be part of the default API?

[dyladan]

I am fine with this. QQ: should this be part of the default API?

Think this discussion is in #428

[bogdandrutu]

@SergeyKanzhelev please review. also @open-telemetry/specs-approvers :)

[dyladan]

The W3C Distributed Tracing WG is discussing this issue this week. Please do not merge this PR yet.

[yurishkuro]

requesting changing to W3C editor draft format

[dyladan]

@tedsuo see w3c/baggage#18

The WG decided not to go with IETF dict, and since this isn't merged yet it isn't too late to change

[dyladan]

@yurishkuro would appreciate a ✅ here

[yurishkuro]

btw, why not otcorreations to match the Correlations API?

[dyladan]

Originally I had been trying to keep the header name as similar as possible to the name I thought the w3c would likely end up with. Because of the recent naming discussions, I think it's clear that we should just use whatever header name makes sense for us.

[dyladan]

Rebased with name otbaggage PTAL and merge when you're happy

[MrAlias]

@dyladan: it seems like the above poll shows otcorrelations as the prefered header name for the time being. Why did we switch to otbaggage?

[dyladan]

@MrAlias oh man I just had a brain fart I guess. updating now.

[dyladan]

Can't see why the build is failing. Just getting a spinner in circle ci that says 0 jobs in the workflow.

[MrAlias]

What I'm seeing:

#!/bin/bash -eo pipefail
make markdown-lint
make enforce-markdown-link-check
markdownlint -c .markdownlint.yaml '**/*.md'
markdown-link-check FAILED => errors:

Run 'make markdown-link-check' to see the errors
Makefile:30: recipe for target 'enforce-markdown-link-check' failed
make: *** [enforce-markdown-link-check] Error 1

Exited with code exit status 2

[MrAlias]

That Makefile command is not really helpful. Running manually:

$ find . -name \*.md -exec markdown-link-check {} 2>&1 \;
...
FILE: ./specification/context/api-propagators.md
[✓] #overview
[✓] #http-text-format
[✓] #fields
[✓] #inject
[✓] #setter
[✓] #set
[✓] #extract
[✓] #getter
[✓] #get
[✓] #composite-propagator
[✓] #global-propagators
[✖] api-correlationcontext.md#serialization

12 links checked.

ERROR: 1 dead links found!
[✖] api-correlationcontext.md#serialization → Status: 400
...

[lizthegrey]

Is "ot" as a prefix potentially confusing when opentracing still exists but is deprecated?

[dyladan]

@lizthegrey would you rather we use otelbaggage? It's only another 2 bytes on the header so I don't think we need to worry too much about payload size. I typically think of ot as OpenTelemetry not OpenTracing, but that's just me and I know not everyone feels that way.

[yurishkuro]

there was never otbaggage in opentracing, so I don't think there's much confusion.

[mwear]

There seems to be some continued confusion about what we're actually changing the header name to. I see discussions about otbaggage, but this PR currently changes it to otcorrelations. I think that is due to #517 (comment).

[dyladan]

There seems to be some continued confusion about what we're actually changing the header name to. I see discussions about otbaggage, but this PR currently changes it to otcorrelations. I think that is due to #517 (comment).

Yes my fault. otcorrelations is the name we decided on in my informal poll, I just can't get baggage out of my head because that's what the WG is going towards.

[dyladan]

Is this waiting on something to merge?

[MrAlias]

@open-telemetry/technical-committee can we merge this so language SIGs have direction on this?