Re: [fetch] Proposed backward-incompatible change: requiring cloning all requests/responses (#61)

@annevk re https://github.com/whatwg/fetch/issues/61#issuecomment-115376269

> a) checks for the offset to be 0

in https://github.com/whatwg/streams/issues/367 and linked comments, the plan was to compare offset at construction time with offset at fetch/cache.put/etc. time, not to check that it's always zero. But I guess that is the subject of https://github.com/whatwg/streams/issues/367#issuecomment-115428031 so for "v1" we will check it's zero?

> Here are some examples to show what this means:

Yay, these are very helpful! This transfer semantic is pretty interesting and I guess does help solve the problem.

Here is one more example that I believe follows from yours:

```js
req = new Request({body: someClosedReadableByteStream})
assert(req.body !== null)
assert_false(req.bodyUsed)
fetch(req) // success
assert(req.body === null) // due to transfer
assert_true(req.bodyUsed) // due to obsolete flag
assert_reject(fetch(req)) // due to obsolete flag
```

i.e., this is equivalent to `{body: ""}`.

---

However, I am afraid of the issue @yutakahirano pointed out in https://github.com/yutakahirano/fetch-with-streams/issues/37#issuecomment-115148298:

```js
req = new Request({ body: "Hello world!" });
req.body.cancel();

// offset is zero since nothing has been read, so
// now this is equivalent to the above case: req.body is just
// a closed, zero-offset RBS.

assert(req.body !== null);
assert_false(req.bodyUsed); // !!
fetch(req);                 // !! success
assert(req.body === null) // due to transfer
assert_true(req.bodyUsed) // due to obsolete flag
assert_reject(fetch(req)) // due to obsolete flag
```

This situation seems fine to me, but I am afraid @wanderview will not like it, since author code managed to disturb the stream in some way and yet fetch still succeeded/bodyUsed is still false.

---
Reply to this email directly or view it on GitHub:
https://github.com/whatwg/fetch/issues/61#issuecomment-117613751