mutating of trace state

[SergeyKanzhelev]

#93

[yurishkuro]

Why MAY mutate instead of SHOULD?

[codefromthecrypt]

+ +Library or platform receiving `traceparent` request header MUST send it to outgoing requests. It MAY mutate the value of this header before passing to outgoing requests.

guess this is for pass-through propagation? If so, this example should be noted I guess.

[SergeyKanzhelev]

Yes, MAY was for the pass-thru. I was thinking to make it more radical and suggest that if traceparent was not mutated - tracestate shouldn't be either. One exception may be when proxy wants to tall something about itself. Like some identifier of a proxy.

Anybody knows real life scenarios when pass-thru for traceparent doesn't mean pass-thru for the tracestate?

[SergeyKanzhelev]

Posted an update. @adriancole @yurishkuro appreciate feedback.

[yurishkuro]

I prefer to separate the players into pass-through only and participating. The former must not mutate. The later must mutate the parent.

The reason MAY bothers me is that it leaves the standard vague. A node can be screwing the overall trace and still be compliant.

[SergeyKanzhelev]

@yurishkuro I added this paragraph. Is it what you think needed?

If the value of traceparent field wasn't changed before propagation - tracestate MUST NOT be modified as well. Unmodified headers propagation is typically implemented in a pass-thru services like proxies. This behavior may also be implemented in a services which currently does not collect distributed tracing information.

[SergeyKanzhelev]

@adriancole I'd appreciate sign off as I can only merge after one contributor approval.

[bogdandrutu]

@SergeyKanzhelev done 👍