Re: [streams] Port precise flow control to writable streams (#318)

> (1) and (2) would work but changed signal is more flexible (yeah, HWM=0 would behave as them). What's the reason you preferred them than changed signal? Simplicity? I understand that the name "ready" isn't good name for "changed" signal, though.

Yeah, basically simplicity. I cannot figure out the use case for the generic changed signal. The generic changed signal seems to not be very aligned with the high water mark idea. That is, I think two valid default strategies are either you should always be writing when desiredSize > 0 (1), or you should wait until the stream is drained and desiredSize = HWM (2). Anything more customizable can be taken care of by specific calls to waitForDesiredSize(n), with a parameter. (Remember, we are just trying to decide the default for `waitForDesiredSize()`, or whatever it is called.)

In fact, maybe it does not matter too much. I think you can get (4) with `writer.waitForDesiredSize(writer.desiredSize + 1)`. (Well, assuming your sizes are integers. If we really wanted to fix that we could do `.waitForDesiredSize(n, { comparison: ">" })` or similar.)

I also noticed when reviewing Node streams that their `'drain'` event is (2), not (1).

Given this I propose: `writer.waitForDesiredSize(n = high water mark)`, i.e. default to (2). This will fulfill when desiredSize becomes >= min(highWaterMark, n).

> Yeah. When HWM of 0 is used, desiredSize would be 0 or negative. The problem is that pipeTo() should be designed to (possibly by having some mode to) be able to push data to such a writable stream.
>
> "Push only when non-negative" would work?
>
> ... Maybe this works?

I think this works if we say waitForDesiredSize waits for >= min(highWaterMark, n). So in this case it will be equivalent to waitForDesiredSize(0), which fulfills as soon as the queue drains.


---
Reply to this email directly or view it on GitHub:
https://github.com/whatwg/streams/issues/318#issuecomment-137286786