Ready for Developer Testing: Future browsing context group dependency hint
Kevin McNee
Contact emails
mc...@chromium.orgExplainer
https://github.com/explainers-by-googlers/future-browsing-context-group-dependency-hintSpecification
NoneSummary
Extends the use of the "opener" rel type to same-window navigations to signal to the browser that the destination page will open a new window and the referring page expects to be able to access it. Some user agents perform browsing context group changes on navigation that aren't strictly necessary for security in order to improve performance. This results in named window reuse not being possible across a back navigation. Annotating an anchor element with "opener" will signal to the user agent that performing a browsing context group change would break a future opener relationship.
Blink component
Blink>Internals>FramesTAG review
NoneNote: See here for the early design review.
TAG review status
PendingRisks
Interoperability and Compatibility
WebKit wouldn't have to make any changes. They already have the intended behaviour of reusing the existing window, without the use of the proposed annotation, and would be aligned with the spec changes. If Firefox wishes to adopt this behaviour, I wouldn't expect architectural difficulties in doing so. It's a matter of passing a bool through to their isolation logic. Alternatively, it would be fine from a spec perspective if they wish to keep the existing behaviour. The use of an opener rel on a link that targets the same window currently has no effect. There are a small number of pages that do this anyway (see use counter [1]). The only negative effect such pages would experience is a potential unintentional loss of bfcache eligibility. Compared to the lack of clarity from creating a new rel type, this seems worth it. For user agents that don't recognize an "opener" window feature, this could change the window selection to a popup instead of a tab. This wouldn't matter for _self navigations. But in any case, a page could explicitly use the feature "popup=0" to avoid this. [1] https://chromestatus.com/metrics/feature/timeline/popularity/4742
Gecko: No signal (https://github.com/mozilla/standards-positions/issues/1059)
WebKit: No signal (https://github.com/WebKit/standards-positions/issues/383)
Web developers: Positive (https://issues.chromium.org/issues/40281878) The reporter of the motivating bug has indicated that this proposal would enable their use case without bfcache-ineligibility hacks.
Other signals:
Security
No concerns expected. This is not introducing a new capability. It's a more targeted opt-out mechanism for behaviour that can already be achieved. Also note that the page can only use this to avoid optional BCG changes that aren't needed for security. Notably, it can't be used to avoid COOP enforcement.
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?
None
Goals for experimentation
To verify that this resolves https://crbug.com/40281878 .Ongoing technical constraints
None
Debuggability
None
Will this feature be supported on all six Blink platforms (Windows, Mac, Linux, ChromeOS, Android, and Android WebView)?
YesIs this feature fully tested by web-platform-tests?
NoFlag name on chrome://flags
NoneFinch feature name
RelOpenerBcgDependencyHintRequires code in //chrome?
FalseTracking bug
https://issues.chromium.org/333743493Estimated milestones
| DevTrial on desktop | 128 |
| DevTrial on Android | 128 |