[horizion2025]

As to the complexity: It is complex because it is very low-level. In JavaScript (using that as an example but I suspect Python is the same) they build in async/await keywords such that they are in cahoots with the Promise class. C++ takes a different path where there isn't a built-in Promise class, rather it provides you lower-level primitive you can use to build a Promise class. You can build a library around it, and it will be as simple as in other languages - both for the awaiter and for implementing libraries that you can await on :) But I agree it is really complicated. I remember once in a while thinking "ah it can't really be that complicated" only to dive into it again. It doesn't help that practically any term they use (promise, awaiter etc.) is used differently than in all other contexts I've worked. If you just expect it will be as easy as understand JavaScript async/await/promise you are in for a rude surprise. Raymond Chen has written co-routines tutorials which span three SERIES. Here's a map of those: https://devblogs.microsoft.com/oldnewthing/20210504-01/?p=10...

As for how we got to here without:

1) Using large number processes/threads 2) Raw callback oriented mechanisms (with all the downsides) 3) Structured async where you pass in lambda's - benefit is you preserve the sequential structure and can have proper error handlign if you stick to the structure. Downside is you are effectively duplicating language facilities in the methods (e.g. .then(), .exception() ). Stack traces are often unreadable. I. 4) Raw use of various callback-oriented mechanisms like epoll and such, with the cost in code readability etc. and/or coupled with custom-written strategies to ease readability (so a subset of #3 really)

With C++ couroutines the benefit is you can write it almost like you usually do (line-by-line sequentially) even though it works asynchronously.