We're working on improvements to our SDK, expanding it from "just an API wrapper" to a more involved part of an application. As the complexity increases, so does the risk of code not working as we'd hoped.

Naturally, we'd like to add telemetry to know what's working, what's not, how often features are used, etc.

To be minimally intrusive, we plan to make it easy to opt out (w/ an obvious callout in the SDK install docs), document exactly what data we collect, and ensure it is 100% anonymous. This aligns with Russ Cox's recent post, "Transparent Telemetry for Open Source Projects" [1].

What else should we keep in mind?

[1]: https://research.swtch.com/telemetry-intro