David Pedersen

@hdevalence sorry this has taken so long to land. My opinion from 2 months still lands. If we fix the small docs things I would be okay with merging this....

I think your proposal sounds good! I'm up for helping investigating the proposed changes and porting tower-http and axum to see how they work in practice. One thing though: It...

> Right, my point is we should use hypers timeline to ensure that we 1.0 for inclusion into hypers 1.0 release. I think it would be pretty important ecosystem wise...

I've filed https://github.com/tower-rs/tower/pull/657 which I think we should at least consider as part of tower-service 1.0

GATs have been right around the corner for a long time 😅 I'm also not sure anyone has done a serious attempt at porting stuff to a GAT based service...

> Related to the split traits approach @olix0r mentions, I remember that a long time ago, `tower` had a `ReadyService` trait that could be implemented by services that are _always_...

I think it's a good pattern to follow. I haven't run into any leaks myself because of it. Rust code tends to be pretty good at freeing resources when they're...

> But dropping is not the issue here, right? The issue is if some shared state is acquired without calling call() it will leak. Tokio'a semaphore releases the permit when...

I've also thought about steer a bit. Not sure its very useful atm due to all the services having to be the same type. Unless users are cool with boxing...

I'm curious what your use case for this is 🤔 I've personally considered removing this blanket impl and instead have a blanket impl for async functions so we wouldn't need...