Keyed vs Direct

[conorbros]

Performance wise, is it preferable to construct a direct RateLimiter for each new connection or use a keyed one?

[antifuchs]

Good question - I'm pretty sure a real authoritative answer will involve benchmarks of your specific code base, but I imagine the following factors will play a role in how easy each method will be to execute:

• do limits apply to the concrete connection or to a higher-order "user" credential? (That is, can one user make multiple connections, and does a single limit apply across all of them?)
• Does your connection allow associating values with it? (like, is it a struct where the limiter could be a field, or is it alive only inside an async fn where you can capture the rate limiter?)

With those in mind, it should be easier to find a representation that'll work for you - if there's a limit per connection (not per user) and you can easily associate a limiter with a connection - make it direct. Otherwise, if you limit per-user, I'd say go with the keyed variant.

Hope that helps - I'll close this issue, but feel free to re-open it you have follow-up questions.