Observing reports (via ReportingObserver) cross-frame

[paulmeyer90]

Should a ReportingObserver registered to one frame be able to observe reports generated in a (same-origin) subframe of that frame? I think I had always envisioned that it should be able to, but would like to hear the thoughts of others on this.

I'm assuming that for the cross-origin case, this kind of observation doesn't make sense, but I'm open to that discussion also!

[patrickkettner]

im not against the concept, but I believe the initial plan was to model after PerformanceObserver, which does not do this. Also, it would be fairly trivial to pass a report up the window chain, right?

[paulmeyer90]

See this bug which was caused by confusion around this issue: https://bugs.chromium.org/p/chromium/issues/detail?id=865779

[paulmeyer90]

As Eric suggested on that bug, I think it would be great to have a new ReportingObserver option (called something like "subframe", "propagate", or "omniscient" :P ?) that sets whether the observer will observe subframes.

[igrigorik]

I'm not against the concept, but I believe the initial plan was to model after PerformanceObserver, which does not do this.

Yep, and I would really like to keep the API surface aligned between the two.

Also, it would be fairly trivial to pass a report up the window chain, right?

Ish. There's a number of edge cases to consider here, which we started hashing out in w3c/performance-timeline#86. If we can land on a solution for the issues covered in that bug, I think we'll have a good solution that can be applied to both observers.

[clelland]

Closing this; it makes the spec simpler to just surface reports in the document where they are triggered; it means that you don't get reports in ReportingObserver that were not also delivered to the enpoints in the Report-To header; and same-site already have ways to share this information if they want to.