Re: [whatwg/streams] ReadableStream constructor: standard text and reference implementation doesn't match (#917)

Looking at https://github.com/ricea/web-platform-tests/blob/f7aab0af8bf4ffc4aa24ce4bb5299b0a063ff8ef/streams/readable-byte-streams/constructor.js#L9 is a bit concerning. The logic is not very transparent. Additionally, it's interesting that we disallow some things but ignore others; e.g. we ignore autoAllocateChunkSize for non-byte streams, but disallow a size() in the strategy.

A variant that would make the most sense to me would be:

```js
  op('get', 'cancel'),
  op('validate', 'cancel'),
  op('get', 'pull'),
  op('validate', 'pull'),
  op('get', 'start'),
  op('validate', 'start'),
  op('get', 'type'),
  op('validate', 'type'),
  op('get', 'autoAllocateChunkSize'),      // *
  op('tonumber', 'autoAllocateChunkSize'), // *
  op('validate', 'autoAllocateChunkSize'), // *
  op('get', 'size'),                       // **
  op('validate', 'size'),                  // **
  op('get', 'highWaterMark'),
  op('tonumber', 'highWaterMark'),
  op('validate', 'highWaterMark'),
```

Then

- * = either omit if type !== 'bytes', or omit the tonumber and change the validate to throw, or just omit the tonumber/validate
- ** = either omit if type === 'bytes', or change the validate to throw, or just omit the validate

The principles here are:

- left to right argument processing
- alphabetical within arguments
- type conversion/validation happens immediately after getting

It's worth noting that the semantics with * = omit the tonumber/validate and ** = omit the validate are then exactly what you'd get with Web IDL.

We'd then want to extend these principles to the other cases besides readable byte stream.

I realize this would be harder to factor, given that it would require logic that is currently in SetUpReadableByteStreamControllerFromUnderlyingSource to be put back in the constructor. So maybe it's not worth it. I'm totally happy to accept #922 and associated tests, and revisit this later if someone else complains. The more important thing is just picking an order, and aligning the reference implementation/tests together, and in that regard #922 suffices. But I thought it'd be good to write out my thoughts.

-- 
You are receiving this because you are subscribed to this thread.
Reply to this email directly or view it on GitHub:
https://github.com/whatwg/streams/issues/917#issuecomment-383617735