Thanks for pointing out the specific spec algos. > @ricea had suggested using an experiment, I assume because the use of fetch > is so widespread and changing the type of error that's thrown could have > consequences, even if it's the intended behavior. But I'll let him comment on > that himself if he has anything to add.
We have generally been recommending against gradual rollout in cases like this: https://chromium.googlesource.com/chromium/src/+/main/docs/flag_guarding_guidelines.md#prefer-waterfall-rollout-for-platform-changes. In short, an experimental rollout makes any breakage hard to track for developers because it won’t show up consistently. The usual approach would be to just ship normally via canary/dev/beta/stable and be ready to roll it back via finch if there are issues. We shouldn’t do either without trying to understand the risk, though. What do you expect potential compatibility issues to look like for this? Do developers often have logic branching on specific abort values for these promises versus just logging them? Do we have Use Counters that could put a high bar on the risk, like a counter for how often already-completed fetches are aborted with a reason, or how often the Response methods are called for a fetch in this state? From: Taylor Brandstetter <[email protected]> Sent: Wednesday, August 12, 2026 2:30 PM To: blink-dev <[email protected]> Cc: Daniel Clark <[email protected]>; Chromestatus <[email protected]>; [email protected] <[email protected]>; Adam Rice <[email protected]> Subject: Re: Intent to Ship: Fetch API: Forward reason from AbortController to fetch Response You don't often get email from [email protected]<mailto:[email protected]>. Learn why this is important<https://aka.ms/LearnAboutSenderIdentification> > Can you please give a more specific link into the part of the spec that > defines this behavior? Sure thing; the body methods point to the steps for consuming a body: https://fetch.spec.whatwg.org/#concept-body-consume-body Which point to the steps for fully reading a body: https://fetch.spec.whatwg.org/#body-fully-read Which say to read all the bytes from the reader. And regarding the reader, the main abort steps (https://fetch.spec.whatwg.org/#http-network-fetch) say: 1. If fetchParams is aborted<https://fetch.spec.whatwg.org/#fetch-params-aborted>, then: * Set response’s aborted flag<https://fetch.spec.whatwg.org/#concept-response-aborted>. * If stream is readable<https://streams.spec.whatwg.org/#readablestream-readable>, then error<https://streams.spec.whatwg.org/#readablestream-error> stream with the result of deserialize a serialized abort reason<https://fetch.spec.whatwg.org/#deserialize-a-serialized-abort-reason> given fetchParams’s controller<https://fetch.spec.whatwg.org/#fetch-params-controller>’s serialized abort reason<https://fetch.spec.whatwg.org/#fetch-controller-serialized-abort-reason> and an implementation-defined<https://infra.spec.whatwg.org/#implementation-defined> realm<https://tc39.es/ecma262/#realm>. So, both the readable stream and all the body methods consuming it should be errored with the abort reason. > Can you request a WebKit signal? I didn't realize it was necessary just for filling a gap in the implementation, but the wide review documentation<https://www.chromium.org/blink/launching-features/wide-review/> does call this case out, so I will request a signal and update the feature later. > I'm still unclear on which tests cover this change. I called them out there; "response.X() rejects with abort reason if already aborted" (for each body method) and "Stream errors once aborted with abort reason". They fail on everything but Firefox (well except for the latter which doesn't run on Firefox); I'm assuming the dashboard wasn't updated when you responded but you can see the failures now. > Can you say more about this roll out plan? This doesn't seem like the kind of > thing that would need to be rolled out gradually via experiment. @ricea had suggested using an experiment, I assume because the use of fetch is so widespread and changing the type of error that's thrown could have consequences, even if it's the intended behavior. But I'll let him comment on that himself if he has anything to add. > Lastly please request the other "Prepare to Ship" reviews in the chromestatus > entry (Privacy, WP Security, etc). Will do. Thanks for your review, Dan. On Monday, August 10, 2026 at 12:11:55 PM UTC-7 Dan Clark wrote: > Specification > https://fetch.spec.whatwg.org<https://fetch.spec.whatwg.org/> Can you please give a more specific link into the part of the spec that defines this behavior? > WebKit: No signal Can you request a WebKit signal? > Is this feature fully tested by > web-platform-tests<https://chromium.googlesource.com/chromium/src/+/main/docs/testing/web_platform_tests.md>? > Yes > https://wpt.fyi/results/fetch/api/abort/general.any.html Specifically the > tests: * response.arrayBuffer() rejects with abort reason if already aborted > (and other response methods such as body()) * Stream errors once aborted with > abort reason. Underlying connection closed. I'm still unclear on which tests cover this change. Looking at the test results<https://wpt.fyi/results/fetch/api/abort/general.any.html?label=master&product=chrome%5Bstable%5D&product=edge%5Bexperimental%5D&product=firefox%5Bexperimental%5D&product=safari%5Bexperimental%5D&aligned> all are already passing in Chrome Stable except for "Readable stream synchronously cancels with AbortError if aborted before reading" which doesn't seem to match this scenario. > Rollout plan > (RARE) Experiment users ramp up over time Can you say more about this roll out plan? This doesn't seem like the kind of thing that would need to be rolled out gradually via experiment. Lastly please request the other "Prepare to Ship" reviews in the chromestatus entry (Privacy, WP Security, etc). Thanks, Dan On Wednesday, August 5, 2026 at 1:46:36 PM UTC-7 Chromestatus wrote: Contact emails [email protected]<mailto:[email protected]> Specification https://fetch.spec.whatwg.org<https://fetch.spec.whatwg.org/> Summary An AbortController can be passed into fetch to allow a request to be aborted; this is already supported see https://chromestatus.com/feature/5631483679080448 When calling abort, you can optionally pass in an "abort reason", and the original fetch promise if it hasn't resolved should be rejected with that reason. This is already working as intended. *However*, if the fetch promise *has* resolved (after reading the header), but the body has not yet been fully read, this is also intended to propagate the abort reason to the Response methods such as Response.blob(), as well as the ReadableStream Response.body. This part is not currently working; the relevant Promises instead are rejected with generic AbortErrors. Firefox at least is compliant here but chromium/Edge/Safari are not. Blink component Blink>Network>FetchAPI<https://issues.chromium.org/issues?q=customfield1222907:%22Blink%3ENetwork%3EFetchAPI%22> Web Feature ID abortable-fetch<https://webstatus.dev/features/abortable-fetch> Motivation No information provided Initial public proposal No information provided TAG review No information provided TAG review status Not applicable Goals for experimentation None Risks Interoperability and Compatibility No information provided Gecko: Shipped/Shipping WebKit: No signal Web developers: No signals Other signals: WebView application risks Does this intent deprecate or change behavior of existing APIs, such that it has potentially high risk for Android WebView-based applications? No information provided Debuggability No information provided Will this feature be supported on all six Blink platforms (Windows, Mac, Linux, ChromeOS, Android, and Android WebView)? Yes Is this feature fully tested by web-platform-tests<https://chromium.googlesource.com/chromium/src/+/main/docs/testing/web_platform_tests.md>? Yes https://wpt.fyi/results/fetch/api/abort/general.any.html Specifically the tests: * response.arrayBuffer() rejects with abort reason if already aborted (and other response methods such as body()) * Stream errors once aborted with abort reason. Underlying connection closed. Flag name on about://flags No information provided Finch feature name ForwardReasonToFetchBodyAbort Rollout plan (RARE) Experiment users ramp up over time Requires code in //chrome? False Tracking bug https://issues.chromium.org/issues/502133195 Estimated milestones Shipping on desktop 153 Shipping on Android 153 Shipping on WebView 153 Shipping on iOS 153 Anticipated spec changes Open questions about a feature may be a source of future web compat or interop issues. Please list open issues (e.g. links to known github issues in the project for the feature specification) whose resolution may introduce web compat/interop risk (e.g., changing to naming or structure of the API in a non-backward-compatible way). No information provided Link to entry on the Chrome Platform Status https://chromestatus.com/feature/5158507786665984?gate=6176253840326656 This intent message was generated by Chrome Platform Status<https://chromestatus.com/>. -- You received this message because you are subscribed to the Google Groups "blink-dev" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To view this discussion visit https://groups.google.com/a/chromium.org/d/msgid/blink-dev/DS0PR00MB2527F68D4E77B90AD4BDCCE8C5DA2%40DS0PR00MB2527.namprd00.prod.outlook.com.
