Host stability badges within otel ecosystem

[lmolkova]

From #278 (comment)

shields.io is designed to provide badges for things like GitHub pipeline status, it's cached by the browser, by the GH and on the shields.io content distribution side.

But it's an external dependency, which might go down or degrade, we might hit rate-limits affecting anyone browsing semconvs.

As an immediate mitigation we can disable badge generation in favor of plain text, but in the long term we should consider hosting badges ourselves:

• on opentelemetry.io
• inside semconv repo
• ...

[arminru]

cc @open-telemetry/docs-maintainers

[cartermp]

I'm not inherently opposed to hosting on the otel.io site, although I think it's more likely that we'd go down than sheilds.io would.

[svrnm]

Are the rate limits the major concern, are we hitting them on a regular basis? Based on the initial comment this is not for build pipelines etc, but for more or less static badges:

        return "![Stable](https://img.shields.io/badge/-stable-lightgreen)"

Our website is hosted on netlify, shields.io on cloudflare, so both should be fairly stable. Hosting it ourselves might add some extra control over caching I guess.

---

Edit: An additional comment, if we choose between any option, then hosting those images within semconv is my least preferred, because with the copy of SemConv (https://opentelemetry.io/docs/specs/semconv/) on the website they would land on opentelemetry.io eventually anyway

[lmolkova]

we're going to have them added to almost every attribute in semconv repo and on otel.io. They are static (and there 3 distinct ones).
Interestingly enough GitHub already caches them plus they are cached by user browser. So in practice I don't expect any rate-limiting problems. If we ever do, we can turn them off - they are not essential. We can also reconsider our approach then.

otel.io (there are multiple badges on the page, but only each badge is requested only once, also cached in memory):

semconv repo (same here)

[svrnm]

yes, looks like caching is solving most of the concerns. And, as long as you keep an alternate title for each badge I don't see any issues even if whatever service is used is down.