LGTM1 Thanks for diving into the samples. Sounds like the breakage risk here is indeed low.
On Wed, Feb 21, 2024 at 8:15 PM Brad Triebwasser <btri...@chromium.org> wrote: > Sure, I checked 10 unique origins (skipping duplicate pu707ev.com > subdomains which make up half of the sampled domains) listed in > https://chromestatus.com/metrics/feature/timeline/popularity/4448. > > 4 of them reference: > https://cdn.jsdelivr.net/gh/cffgnu/qhdd/asset/responsive.min.js. I won't > post all the code here, but you can see it querying a list of permissions, > with a .catch() shortly after. Evidence of fingerprinting, but does handle > exceptions. > 1 on the root page <https://app.webscrapingapi.com/login> listed all > permissions (and specifically > referenced > third_party/blink/renderer/modules/permissions/permission_descriptor.idl); > Looks like fingerprinting. It has a catch following the > navigator.permissions.query call. > 1 had a library > <https://klempner-verband.de/assets/js/app.js?id=a58e3fa4bf16bb1db6a6.js>checking > every permission with navigator.permissions.query with a corresponding > catch handler. > 2 referenced https://fs.pudaf.com/fp.js. This was a little more > obfuscated and I can't tell if exceptions are handled, but looking at the > code, there is a ton of evidence suggesting fingerprinting (iterating over > every permission, navigator properties, etc.). I tried in Firefox, and > didn't observe any unhandled exceptions even though the debugger did pass > that point in the code. > 2 used a similar highly obfuscated library > <https://ebarter.pro/app/js/dd.js> which contains the "window-placement > string" but I did not see any corresponding permissions.query call, so > results are inconclusive, but the sites did load in firefox with no > unhandled exceptions on the missing window-placement permission. > > Out of 10 sampled origins, 8 were obvious fingerprinting, 8 had exception > handling, 2 were too obfuscated to be conclusive. > > Regards, > Brad > > On Tue, Feb 20, 2024 at 11:01 PM Yoav Weiss (@Shopify) < > yoavwe...@chromium.org> wrote: > >> >> >> On Tuesday, February 20, 2024 at 8:32:33 PM UTC+1 Brad Triebwasser wrote: >> >> [Premature Send. Full message below] >> >> We are tracking UMA for when the permission name "window-placement" is >> parsed (e.g all calls to navigator.permissions.query({name: >> 'window-placement'}) increment the counter. That function will throw an >> exception *"The provided value 'window-placement' is not a valid enum >> value of type PermissionName." *once we remove the permission. So based >> on the metrics, 0.006% of page loads that aren't handling the exception >> could break. I strongly suspect most sites would have exception handling >> since no other browser has implemented this permission string. >> >> >> Would it be possible to manually inspect a few samples to see how many of >> these sites properly handle the exception? >> https://chromestatus.com/metrics/feature/timeline/popularity/4448 seems >> to have a list of 100+ origins >> >> >> We are also tracking UMA for window-placement permission policy. So >> everytime the browser parcels "window-placement" in a header (e.g. >> Permissions-Policy: >> window-placement=(self)) or in an iframe (e.g. <iframe src=" >> https://example.com"; allow="window-placement"></iframe>), the counter is >> incremented. So ~0.015% of page loads are parsing the window-placement >> policy. This scenario would not explicitly break a page, but the policy >> would silently be ignored and the corresponding permission denied if the >> site did not also have window-management specified. Again, no other browser >> has implemented this string, so I suspect sites legitimately using this >> would have some kind of fallback for non-chromium browsers anyway. >> >> Regards, >> Brad >> >> >> On Tue, Feb 20, 2024 at 11:24 AM Brad Triebwasser <btri...@chromium.org> >> wrote: >> >> Certainly. >> >> We are tracking UMA both for when the permission name "window-placement" >> is parsed (e.g all calls to navigator.permissions.query({name: >> 'window-placement'}) increment the counter. That function will throw an >> exception *"The provided value 'window-placement' is not a valid enum >> value of type PermissionName."* >> >> On Mon, Feb 19, 2024 at 12:10 PM Yoav Weiss (@Shopify) < >> yoavwe...@chromium.org> wrote: >> >> >> >> On Fri, Feb 16, 2024 at 8:19 PM Brad Triebwasser <btri...@chromium.org> >> wrote: >> >> Usage for the legacy permission and permission policy are ~0.006 >> <https://chromestatus.com/metrics/feature/timeline/popularity/4448> and >> ~0.015 >> <https://chromestatus.com/metrics/feature/timeline/popularity/4450> >> >> >> Can you detail what these two different counters represent? >> Our typical threshold is about half of the lower one (~0.0003%), but that >> varies based on the potential breakage. >> >> What would breakage here look like? >> >> >> (% page loads) while the new variants are ~1.166 >> <https://chromestatus.com/metrics/feature/timeline/popularity/4447> and >> ~3.066 >> <https://chromestatus.com/metrics/feature/timeline/popularity/4449> (% >> page loads) respectively, so the main metric for potential breakage here is >> 0.006% of page loads. 0.006% seems low in my opinion, but I'm curious if >> there is any guidance on a lower target % we should be aiming for prior to >> removing this feature. >> >> I separately calculated that the 200 origins make up about 0.0058% of >> total origins tracked by UKM (if that's what you're asking), which aligns >> with the UMA figures. I also want to note that 50 of those origins are >> likely the same (spam?) website (e.g. https://aaaa123.com, >> https://aaaa124.com, https://aaaa125.com). You can see examples of this >> in the UMA sample data >> <https://chromestatus.com/metrics/feature/timeline/popularity/4448>. >> >> Regards, >> Brad >> >> On Thu, Feb 15, 2024 at 12:42 AM Yoav Weiss (@Shopify) < >> yoavwe...@chromium.org> wrote: >> >> 200 origins sounds like a lot. Do you know what %age of page views those >> origins would represent? >> >> On Wed, Feb 14, 2024 at 11:32 PM Brad Triebwasser <btri...@chromium.org> >> wrote: >> >> *tl;dr:* We expect at most 200 origins could break, and only ~30 of >> those may be legitimately using the API. >> >> We do track UMA/UKM for the primary API entrypoint function ( >> GetScreenDetails >> <https://developer.mozilla.org/en-US/docs/Web/API/Window/getScreenDetails>) >> which we expect nearly all legitimate usage of the API to use. We see 60 >> unique origins invoking GetScreenDetails, dominated by a handful of >> origins (which align with the partners we know are using the API). >> >> For reference, 2500 unique origins are checking the window-management >> permission, and 200 unique origins checking the old window-placement >> permission (82% of those origins are *not *logging any GetScreenDetails >> calls). >> >> As Mike mentioned, the only breakage here would be a site using >> navigator.permissions.query({name: 'window-placement'}) without error >> handling which according to UKM data would be roughly 200 origins (at most >> 18% of those may be legitimately using the API). >> >> I believe 200 unique origins is a relatively low number of potential >> breakages, especially considering our data strongly suggests a majority of >> that is fingerprinting. >> >> Regards, >> Brad >> >> >> On Mon, Feb 12, 2024 at 6:27 AM Mike Taylor <miketa...@chromium.org> >> wrote: >> >> Agree that the risk feels low... one thing to perhaps check for (if you >> have UKM or use counters) is to see if there is any legit usage on sites of >> `navigator.permissions.query()` that isn't catching errors, since that >> will throw a TypeError and can break a page. >> On 2/12/24 9:16 AM, Rick Byers wrote: >> >> Presumably the risk of legitimate breakage here is bounded by the use of >> the Window Management API, right? Are there any UseCounters for the various >> Window Management operations? I couldn't find any at a quick glance. I >> imagine legitimate usage is dominated by a few sites with an obvious need >> (do we have UKM data?), and such sites should always degrade gracefully >> without window management capabilities, right? >> >> My intuition is that the compat risk here should be extremely low, but I >> hope we have some data to validate that which isn't tainted by the >> fingerprinting usage. >> >> Rick >> >> >> On Wed, Feb 7, 2024 at 1:59 PM Brad Triebwasser <btri...@chromium.org> >> wrote: >> >> +blink-dev@chromium.org <blink-dev@chromium.org> / Reply All >> >> Thanks for your feedback, Mike! Recipes inline: >> >> On Tue, Feb 6, 2024 at 9:36 PM Mike Taylor <miketa...@chromium.org> >> wrote: >> >> Hi Brad, >> On 2/6/24 3:49 PM, Brad Triebwasser wrote: >> >> Contact emails >> >> btri...@chromium.org >> >> Explainer >> >> https://github.com/w3c/window-management/blob/main/ >> EXPLAINER_spec_and_permission_rename.md >> >> Specification >> >> https://w3c.github.io/window-management/#api-permission-api-integration >> >> Summary >> >> Removes the legacy "window-placement" alias for permission and permission >> policy "window-management". This is a follow-up to >> https://chromestatus.com/feature/5146352391028736 and corresponding >> blink-dev PSA >> <https://groups.google.com/a/chromium.org/g/blink-dev/c/Hf2b1-S39Uw/m/YAEC_0DSBQAJ>. >> The "window-placement" alias has been showing console deprecation warnings >> since M113 >> <https://chromium.googlesource.com/chromium/src.git/+/13204be718225ae09c8ba7e36b055a369c36c878>. >> We will disable WindowPlacementPermissionAlias >> <https://source.chromium.org/chromium/chromium/src/+/main:third_party/blink/renderer/platform/runtime_enabled_features.json5?q=-f:gen%2F%20AND%20-f:out%2F%20WindowPlacementPermissionAlias> >> by default, and remove the flag and legacy code shortly thereafter. >> >> I'm a little bit confused here - it seems like the PSA of the alias is >> being treated as the beginning of a deprecation, is that correct? My >> interpretation of "will lead to a deprecation and removal" from the >> original message was that it would be followed with an Intent to Deprecate >> and Remove (per https://www.chromium.org/blink/launching-features/# >> deprecate), but it seems like that step of the process was skipped. >> >> Yes, I never sent out a separate "Intent to Deprecate" in this case. The >> original PSA >> <https://groups.google.com/a/chromium.org/g/blink-dev/c/Hf2b1-S39Uw/m/YAEC_0DSBQAJ> >> was >> intended to be a hybrid of the introduction of the new names and >> deprecation of the old ones so we also landed deprecation code (DevTools >> deprecation warnings etc.) during that time. Since these have already been >> "deprecated" since M113, I wasn't sure if a separate "intent to deprecate" >> was appropriate in this case since we already deprecated them and monitored >> usage to be sufficiently low, but I can back-up and send an I2D if >> recommended here. >> >> >> Blink component >> >> Blink>Screen>MultiScreen >> <https://bugs.chromium.org/p/chromium/issues/list?q=component:Blink%3EScreen%3EMultiScreen> >> >> TAG review >> >> No feedback was specifically requested for the permission rename, >> however related TAG reviews have been requested with both the old (1 >> <https://github.com/w3ctag/design-reviews/issues/413>, 2 >> <https://github.com/w3ctag/design-reviews/issues/602>) and new >> terminology (3 <https://github.com/w3ctag/design-reviews/issues/840>). >> >> TAG review status >> >> Not applicable >> >> Risks >> Interoperability and Compatibility >> >> There are low compatibility risks. Usage for the legacy permission and >> permission policy are ~0.006 >> <https://chromestatus.com/metrics/feature/timeline/popularity/4448> and >> ~0.015 >> <https://chromestatus.com/metrics/feature/timeline/popularity/4450> (% >> page loads) while the new variants are ~1.166 >> <https://chromestatus.com/metrics/feature/timeline/popularity/4447> and >> ~3.066 >> <https://chromestatus.com/metrics/feature/timeline/popularity/4449> (% >> page loads) respectively, indicating most usage has already migrated. >> >> These percentages are still relatively high, especially for the >> permissions policy variant. Besides the obvious fingerprint.js usage (which >> shouldn't break pages... I would hope), can you describe what the failure >> mode is after the proposed removal is? Have you dug into the remaining >> usage to verify? >> Yes, I dug into the remaining usage quite extensively via Web Archive >> queries and UKM and couldn't find any usages other than what looked like >> fingerprinting. After removal, the permission API will produce an error due >> to an unknown permission, and the permission policy will silently fail >> (e.g. iframes with allow='window-placement' will not have access to the >> features). I beleive that the numbers shifting several orders of magnitude >> in favor of the new strings seems to indicate legitamite usage has >> migrated, and the remainig usage likely fingerprinting. >> >> Gecko: No signal >> >> Firefox has not implemented the API and corresponding permission yet. The >> original API signal request is here >> <https://github.com/mozilla/standards-positions/issues/542>. >> >> WebKit: No signal >> >> Safari has not implemented the API and corresponding permission yet. The >> original API signal request is here >> <https://github.com/WebKit/standards-positions/issues/117>. >> >> Mind linking to the original API position requests here in this thread? >> Added links above to the original API signal request. FWIW, we have since >> filed additional requests for functionality related to window management, >> not necessarily window *placement* related (hence motivation for >> renaming the API): eg 1 >> <https://github.com/WebKit/standards-positions/issues/96> 2 >> <https://github.com/mozilla/standards-positions/issues/712> >> >> >> Web developers: We have communicated internally with partners using the >> API who have expressed commitment to updating the permission strings in >> their code. >> >> Other signals: Positive comment >> <https://github.com/w3c/window-placement/pull/115#pullrequestreview-1159676614> >> from W3C WG Chair >> >> WebView application risks >> >> This is considered low risk. It removes an alias without any change in >> behavior of the underlying API. >> >> Does this permission do anything on WebView? I would have guessed no. >> Your correct, this window management API doesn't apply to WebView so >> there is no impact there. >> >> >> Debuggability >> >> Disabling WindowPlacementPermissionAlias >> <https://source.chromium.org/chromium/chromium/src/+/main:third_party/blink/renderer/platform/runtime_enabled_features.json5?q=-f:gen%2F%20AND%20-f:out%2F%20WindowPlacementPermissionAlias> >> will stop DevTools deprecation warnings for usage of the legacy strings and >> instead will act as if they did not exist at all (e.g. Permission API will >> produce an error when using "window-placement"). >> >> >> Will this feature be supported on all six Blink platforms (Windows, Mac, >> Linux, ChromeOS, Android, and Android WebView)? >> >> No. This feature is not supported on Android. >> >> Is this feature fully tested by web-platform-tests >> <https://chromium.googlesource.com/chromium/src/+/main/docs/testing/web_platform_tests.md> >> ? >> >> Yes. Web Platform tests have already been migrated to the new alias: >> >> https://github.com/web-platform-tests/wpt/tree/master/window-management >> >> Flag name on chrome://flags >> >> None >> >> Finch feature name >> >> WindowPlacementPermissionAlias >> >> Requires code in //chrome? >> >> False >> >> Tracking bug >> >> https://bugs.chromium.org/p/chromium/issues/detail?id=1328581 >> >> Estimated milestones >> >> M123 (flag disable) M125 (flag/code removal) >> >> Anticipated spec changes >> >> None >> >> Link to entry on the Chrome Platform Status >> >> https://chromestatus.com/feature/5137018030391296 >> >> 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 blink-dev+unsubscr...@chromium.org. >> To view this discussion on the web visit https://groups.google.com/a/ >> chromium.org/d/msgid/blink-dev/CALEeEUCdqsmmEhBROkinxbzTULFPX >> nC8goANs6-_O8n3%2B%3D47hQ%40mail.gmail.com >> <https://groups.google.com/a/chromium.org/d/msgid/blink-dev/CALEeEUCdqsmmEhBROkinxbzTULFPXnC8goANs6-_O8n3%2B%3D47hQ%40mail.gmail.com?utm_medium=email&utm_source=footer> >> . >> >> -- >> 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 blink-dev+unsubscr...@chromium.org. >> To view this discussion on the web visit https://groups.google.com/a/ >> chromium.org/d/msgid/blink-dev/CALEeEUBYrF50-%3Dp8umAxQLaEttR- >> jW4WRfWyF5AATV2p29w17w%40mail.gmail.com >> <https://groups.google.com/a/chromium.org/d/msgid/blink-dev/CALEeEUBYrF50-%3Dp8umAxQLaEttR-jW4WRfWyF5AATV2p29w17w%40mail.gmail.com?utm_medium=email&utm_source=footer> >> . >> >> -- >> 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 blink-dev+unsubscr...@chromium.org. >> To view this discussion on the web visit https://groups.google.com/a/ >> chromium.org/d/msgid/blink-dev/CALEeEUDd9YoriLSXG3h7usjpyJThZ >> 4W3%2BixTCdVK2PhS0p9_Rw%40mail.gmail.com >> <https://groups.google.com/a/chromium.org/d/msgid/blink-dev/CALEeEUDd9YoriLSXG3h7usjpyJThZ4W3%2BixTCdVK2PhS0p9_Rw%40mail.gmail.com?utm_medium=email&utm_source=footer> >> . >> >> -- 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 blink-dev+unsubscr...@chromium.org. To view this discussion on the web visit https://groups.google.com/a/chromium.org/d/msgid/blink-dev/CAOmohS%2BkYoJShdaKKonG1pZasT4vg7enXWdh%2BU8FNeVTmyPKWw%40mail.gmail.com.
