Franz Busch

> So the one thing that mustn't happen is that you pass your scope's `}` and still have stuff running. Very clearly, that's exactly what `execute` is currently doing. I...

> Instead, I think the right model is an explicit ownership transfer from connection pool to child task when a connection is required. And from child task back to connection...

> Wait, wait, who would want to adopt `~Escapable` for connections? That doesn't make any sense to me. That makes them useless. > > non-escaping (probably) makes sense for stuff...

I thought about this some more on the weekend and remembered why so far I decided against scoped ownership transfer. It was due to how cancellation works. The problem that...

I agree that this is something the underlying async sequence can change; however, I don't think you can wrap it. The problem is if the user's task gets cancelled and...

> I think it's possible by wrapping (such that you get to make the cancel in `next` call hand back a connection instead of closing it) but I agree that...

I agree. We **must never** import `CoreFoundation` on non-Darwin platforms. We don't guarantee any stability around that and with the re-core `CoreFoundation` is going away long term on non-Darwin platforms

At this point we cannot adopt `~Copyable` for `Service`s since we need move the service into a child task. Until we get call-once closures that's not going to work nicely....

Yes that's the recommend pattern to check if it exists and import only that. IMO that's the way we should do it here and move the extension for URL to...

Thanks for tackling this. I and the rest of the team agree that this is a problem but we are not yet sure how we want to fix it. There...