Eliza Weisman

Do we want to address this for 0.5?

Aghh! This is definitely my fault, whoops. This is _technically_ a breaking change, but I think it might be semver-compliant to fix it in a minor version, since the previous...

I'm happy to work on implementing this, but I suspect @carllerche will have some opinions on how we ought to do it.

> My original thought was to not do backoff in the reconnect middleware, instead keep back off as part of `Retry` (which doesn't exist yet) and then `Retry` would provide...

A question that came up up while working on this is: what timer should be used for the backoffs? In Conduit, I've been working on an abstraction over timer implementations...

> Am I going crazy but can't you just do this with `Buffer`? Right now the `tokio::spawn` is hard coded https://docs.rs/tower/0.4.8/src/tower/buffer/service.rs.html#76 but in theory you could use https://docs.rs/tokio/1.10.1/tokio/runtime/struct.Handle.html#method.enter maybe or...

Hmm, I wonder if this can be implemented in a more modular way, outside of `buffer`, so that users who don't need to use this don't have to pay for...

i would be fine with a PR to add something like that in `tower`, although i think it should be fairly straightforward for users to implement themselves. out of curiosity,...

> There seems to be one subtle difference between `ServiceFn` and blanket impl for async fns. `ServiceFn` has unconstrained `Debug` impl while async fns and closures returning futures do not...

I did some additional poking around this morning, modifying existing TLS integration tests to use `TrustCredentials.Unspecified`, and it appears that this setting really just does not work inside of ScalaTest....