Michael P. Jung

> I've thrown together an implementation of the concurrency primitive which I think will solve the problem here: (...) I've skimmed over the implementation and the thing I wonder most...

> You need to track writer handles because you need to know which resource is currently being written to, and you need to move it to the freed resources when...

The `PoolBuilder` should probably be replaced by a `ConfigBuilder` that creates immutable `Config` structs. That way the `Pool` could just provide a `set_config` method that takes a `Config` object and...

Neither `bb8` nor `r2d2` support changing the configuration at runtime. That's why I want to replace the `PoolBuilder` by a `ConfigBuilder` and add a `Pool::set_config` method that can be used...

I want to use the [arc-swap](https://crates.io/crates/arc-swap) crate to implement that feature. It should be as simple as putting both the `config` and `manager` into an `ArcSwap`. This will be a...

After some consideration I feel like the manager should not be replaced but the managers themselves should provide a way to modify the internal configuration instead. e.g. the `deadpool-postgres` manager...

This is something that's very similar to something I'm thinking about for `deadpool-postgres`. Instead of creating multiple connections to the DB it's possible to use pipelining. It's quite similar to...

I'm currently tinkering with #196 which should allow to implement this feature. The most difficult part about it is the detection and handling of broken connections. That part could be...

`CreatePoolError::Build` is used for one thing right now. When you specify timeouts and don't specify a `Runtime` the building fails with `CreatePoolError::Build(BuildError::NoRuntimeSpecified)`. I could get rid of that but every...

As far as I understand the marker trait the core of deadpool should mostly be `UnwindSafe`. I have to double check the `Drop` implementations, but if I'm not mistaken the...