Eliza Weisman

> The only nice way around this that I can think of is to _also_ allow the p2c load balancer's random picking to be weighted, probably by using [weighted random...

Doing a bit of additional research, it seems like there might be some methods for generating weighted samplings in _O_(1) time by pre-computing probability tables, as described [here](https://www.keithschwarz.com/darts-dice-coins/). If we're...

ugh, i guess the bans are failing because the `url` patch updates the `idna` dep major version and `trust-dns` depends on `idna` 2.x. should we allow that?

I'm happy to put together a quick PR for this, but I figured it was worth opening an issue first.

> Why not let users downcast it? If http will also 1.0 I don't see the issue in that? Otherwise, I agree with what you said. When `http` 1.0 is...

@seanmonstar In that case, it might be nicer to expose `http`'s errors as a `source` and `downcast_ref` target!

cc @carllerche

I'm open to this, although I was initially pretty leery about printing anything related to span events by default since I feel like it has the potential to get a...

> This happens because the current approach simply uses the `Debug` implementation to record the value: https://github.com/xd009642/tracing/blob/062a6ea6ec371c26d9fa8fa7b1651a5c41f9e67e/tracing-core/src/field.rs#L193 but for this we don't even need any `valuable` integration. And this seems...

The reason this method takes arrays rather than slices is in order to validate that their lengths are less than 32 elements statically. This is used to emit an error...