https://github.com/zenparsing/async-iteration/ is where async iterable is proposed.

The difference is that with an observable the creator of the observable defines the pace at which "events" are produced (push), whereas with an async iterable, the consumer defines the pace at which work is performed. The latter is a much better match for filesystem APIs, e.g. Windows's FindFirstFileEx/FindNextFile/CloseFind, or POSIX opendir/readdir. In these cases the async iterable's .next() would map pretty directly to FindFirstFileEx/FindNextFile and readdir.

/cc @zenparsing