zzydxm

Added second commit to illustrate the change in ssl application

Pushed a new version that fixes the first application data issue by always using {active, false} slides: https://docs.google.com/presentation/d/1V0hmWCBzuSYJA4Yh-SC8WOnf1_k1ZqKdEqhkcANrAss/edit?usp=sharing discussion doc: https://docs.google.com/document/d/1psr-MEmYQHsZ6db8oM8ZPMqveHQBvZ6SeFZLZaTT26Q/edit

Thanks for all the comments! We do have a single dist protocol module file achieves kTLS without OTP modification and is running on all nodes we have. I will clean...

Example dist module here, will do a more complete version in a different PR: terminal 1: ``` erl -proto_dist ktls -sname a@localhost ``` terminal 2: ``` erl -proto_dist ktls -sname...

> But about early data I think it does not have to be a problem for the distribution case as then we have control over the clients and as long...

I think there is an one-line change in ssl application that can make ktls_dist work without timer:sleep It is to set {active, false} in client tls_handshake_1_3:do_wait_finished Then, in server ktls_dist:set_ktls...

I did a new commit which simplified a lot in ssl application: https://github.com/zzydxm/otp/commit/018285e0184be9a5337fa2f41d95de91c3c0f82c to use it: terminal 1: ``` erl -pa ./ -proto_dist ktls -sname a@localhost (a@localhost)1> ktls_dist:r(). true ```...

A different approach with patch in inet_tls_dist instead of a separate ktls_dist module: https://github.com/zzydxm/otp/commit/ab34555210a994083de8f7916e9c0605728f7554

I think the support of process dist control is relative straight forward. Almost all operations can be easily migrated to use socket nif + process dist control, just with a...

> The patch of the `inet_tls_dist` module [zzydxm@ab34555](https://github.com/zzydxm/otp/commit/ab34555210a994083de8f7916e9c0605728f7554) has got a nice size and complexity. But any module in the SSL application should be prefixed with `ssl_`, and also, I...