Ben Darnell

In HTTP/1, the spec is clear that headers are case-insensitive, and mixed-case headers are the norm. We can't change this to satisfy applications that mistakenly treat headers as case-sensitive (I've...

I'm still not sure whether a change like this would be worth the disruption, but I'm sure it's not worth an option. Either leave it as it is and lowercase...

There's no interoperability issue, and the just-released new HTTP RFCs help explain why. The `HTTPHeaders` class is a data structure representing "Fields" as defined in [section 5 of RFC 9110](https://httpwg.org/specs/rfc9110.html#fields),...

Alright, let's start over. Tornado does not include any HTTP/2 support, and I think the talk of HTTP/2 is a red herring. You've got an HTTP/1.1 server that is handling...

I think that restructuring the websocket loop to use two separate tasks (one for control frames and one for user messages) makes sense. But there would still be some limit...

When `fork_processes()` returns a child ID, you're inside that child process. `fork_processes()` never actually returns in the parent process (except by raising an exception). The parent process blocks and handles...

Ah, that makes sense. In that case, I think making the children dict a global would be a simple solution. It's fine to make it a global because `fork_processes` is...

> It's possible to double-fork and I am doing this in one use case (i.e. call fork_processes(2, 0) and inside the childs call fork_processes(0, 0)). Yes, but in that case...

> However, .add_callback is useful only for .put and .put_nowait methods, and not for .get and .get_nowait. There's no way to get the result back from .add_callback. That's true, but...

I think this may be related to covariance (see https://mypy.readthedocs.io/en/stable/generics.html#variance-of-generic-types), like #3119. I think some uses of Dict in type annotations may need to be changed to Mapping (just like...