Exposition format: Versioning

[beorn7]

We are still stuck with v0.0.4. With maturing the exposition format, we should bump the version to utlimately 1.0.0. Also, we need to figure out how to document older versions, and what requirements we have for support of various versions by implementers.

[beorn7]

We should also get clarity where and how to advertise/mark the version, cf. prometheus/common#34

[brian-brazil]

The way I'm thinking is that a 1.0.0 parser would also handle 0.0.4 output, with no more than minor mangling (e.g. the comment escaping). The only real problem with that is if we introduce the end marker.

[grobie]

On a different note, it'd be helpful to include information about Prometheus in the Content-Type header value. text/plain; version=0.0.4 doesn't help to identify the format definition if someone doesn't know about Prometheus.

[RichiH]

That brings up a cyclic point... It would be best to refer to the RFC in there, but we can't do that for the year or so it takes to get to the RFC... A backlink we commit to hold accessible forevermore which can then point to the specs? A possible way out would be to specify "MUST link to $url OR specify RFC $tbd" Richard Sent by mobile; excuse my brevity.

[hdost]

https://github.com/OpenObservability/OpenMetrics seems to make this irrelevant.