CppCon 2018: R. Leahy “The Networking TS in Practice: Testable, Composable Asynchronous I/O in C++”

[@bimmerboard]

FANTASTIC presentation. At first, I thought he was going to be a bit stiff, but it soon became evident that this guy is all business and serious about teaching the audience. It was great that he didn't allow questions to slow down the presentation or take it off topic, as so many others do. I loved it. Probably one of the all-time best CppCon topics.

[@Dar1gaaz]

I do enjoy his style of speaking. Makes it easy to follow and pay attention

[@mrlithium69]

Very useful and practical and good speaker.

[@iddn]

Great talk. Can't wait to try this out in 5 years!

[@benjamin.burton]

4 more to go.

[@kianemami9090]

@@benjamin.burton 2 more to go

[@mangmaia]

👀

[@Radioguy00]

Excellent presentation. Thanks

[@CppCon]

Glad you enjoyed it!

[@etorrie]

So much content in this talk. Need to go back and rewatch a couple more times.

[@simfoo]

The more recordings I watch, the more I regret that I was not at the conference. Great talk!

[@norikazuoshiro6324]

nice talk, really learn something.

[@zhaoli2984]

It would be nice to extend the IO model to other IO types.

[@OperationDarkside]

As awesome as the presentation was, this talk confused me more, than what I already knew about the Networking TS. It's own style of thinking. I'm used to synchronous or concurrent, but this is a model I yet have to grasp.

[@jaysistar2711]

The Coroutines TS will allow simplification of this model to be visually simmilar to syncronous code. I'll be much happier when that's available.

[@loomismeister]

I've been testing similar code with the standalone asio, and I do think there is a bit of hand waving that happened during the mock_async_stream section for testing the async code. Robert correctly notes that the completion handler for the stream won't be called until the io_context is polled, but the synchronous code inside the write_some function is executed as soon as async_write is called on that stream. One might have expected that you could chain multiple async_writes together, with none of the buffers being written to until the io_context is polled, but really they are executing immediately. The fact that an io_context is used for the executor doesn't actually prevent work from being done until the io_context is polled.

[@lefteriseleftheriades7381]

I don't get why async_wait_then_write is a free function and not a member function. Can someone please explain?

[@TheBlenderblob]

not gonna make it until 2023 btw

[@jozoppi]

it is too complex. When i told him that it was too complex , he answered with flexibility.

[@beaujamo]

I created a reactor asio (no boost only pure standard cpp17) from scratch 2 years ago and proactor pattern 1 year ago with working http, tls and other protocols pending support. Works on mac, windows and linux and uses cpp17. Curious if somebody will be interested to hire me for this? Thank you for reading