ZOt6

> Discussing internally: though it would be a breaking change, we thought it would be preferable to make the existing `cleanup()` an async function for simplicity, especially while we are...

I will try this approach, but maybe the catch is to modify various interfaces. Such as: ``` fn write(bytes: &[u8]) ``` into: ``` fn write_vectored(bytes: &[u8]) ``` and ``` fn...

> I have better idea. > > we don't need intermediate vector - and you can use BufWriter which has tons better performance > > impl for Message: > >...

Seems that refactor to `async fn load(&self) -> Result;` would be better?

> Yes, this sounds good, ideally, the storage would just return the data and do not expect mutable struct passed into it to change its internal state. Most of the...