Hugo Landau
Hugo Landau
> > If an application wants to do its own polling, it can do so. It would do so by determining the underlying network resources using `SSL_get_[rw]poll_descriptor` and registering them...
> So I would imagine that an application would like to write it's main loop so that it calls a single function, and then gets a work item. A work...
> That seems suboptimal to me. Which is why the registered-mode API is proposed in this document. `SSL_poll` is just for simple and low-radix use cases. >I would expect some...
> It's all about using the same queue as epoll already gives you. The registered-mode API will obviously have an epoll implementation on the backend for Linux.
> I think a callback will make it a lot easier for applications to adopt. Please elaborate. It's an explicit objective of our QUIC API design efforts to minimise the...
>I suggest we have a queue of events per quic connection. So after calling a function that did something, you just check the queue for events, but only for that...
@kroeckx This is a valuable discussion so I've moved it to a separate issue: https://github.com/openssl/openssl/issues/24125 I think this PR is ready. Rebased and updated.
Merged to master, 3.0 and 3.1. Thank you.
Updated to fix CI.