Thank you for your quick reply Yoav, Please find my answers inline.
On Wednesday, March 29, 2023 at 4:35:32 PM UTC+9 Yoav Weiss wrote: Thank you so much, Javier! :) That's some great analysis! On Wed, Mar 29, 2023 at 7:51 AM Javier Garcia Visiedo <[email protected]> wrote: Hi all, Please find the summary of my findings, after analyzing the UKM data. Currently, the UKM data shows 2,087 distinct domains (eTLD+1) sending a wildcard for ACAH and/or ACAO in response to a credentialled request. The UKM data is not good at showing events over time, but the use counter <https://chromestatus.com/metrics/feature/timeline/popularity/3873> indicates the feature use is currently at 0.12% of total page loads, and the monthly average of page loads has been steadily increasing since the introduction of the use counter. Out of the 2,087 domains found in the UKM, the top one accounts for 39.24% of the total navigations in the UKM (54% within top 20 sites), using this feature. The user experience is not visibly impacted for this site, as CORS is blocking calls to a publisher.Ads are displayed properly, so apparently the blocked calls are impacting some kind of targeted ads API. This site came into the data set for the first time in March 2023, and it’s very popular, so we expect to observe a big bump on the use counter from now on. This site is an outlier in terms of traffic volume, compared to other sites present in the UKM, and given the lack of Ux impact we decided to exclude it from the analysis presented below, to get a better representation for the rest of the sites. Excluding the top site, we tested the next top 20 domains in number of navigations using this feature, accounting for 42% of the total navigation in the UKM. In addition to this, we sampled the long tail for 11 more sites randomly, for an additional 3,4% of the total navigation in the UKM. Extrapolating data from the sampling, we classified sites in the following categories: - No Ux impact: 94.85% of the navigations are free from visible user experience impacts. In most cases, CORS policy is blocking calls to personalization or analytics APIs. "CORS policy" here refers to the status quo before this feature ships? No, these calls are blocked when activating the feature - Impact to critical functionality, overall experience, log-in, etc.: 2.6% - Impact to ancillary elements being blocked by CORS policy, such as ads or widgets not essential to the overall page functionality: 2.55% Putting these numbers in context of the total page loads reported by the use counter, 0,0062% of the total page loads experience some UX regression due to the introduction of this feature. Of these, 0,0031% was unusable. We reached out to the site owners with Ux impacts, and we were able to confirm that 100% of the sites identified to have critical impact have already migrated out of this feature. That's a bit higher than what we're typically comfortable with, but it's great to hear that sites you reach out to react quickly. How much of that impact was in the top 20 (where we can reasonably reach out) vs. the long tail (where it'd be significantly harder to communicate things beyond deprecation logs and reports)? For sites with high impact (already addressed) 100% were found in the top 20. For sites where some visual elements are blocked, 55% (1.48% of total navigations registered in the UKM for this feature) are part of the top 20, 45% in the long tail, with 1.07% of total navigations in the UKM. Happy to further clarify the data if needed. Thanks Javier On Wednesday, July 13, 2022 at 2:49:30 PM UTC+9 Yoav Weiss wrote: Hey Javier! The benefits of UKM is that it can give us a list of URLs that have some breakage potential. The laternative is to cross the usecounter with HTTPArchive data, but that has a strong bias towards homepages, so may miss a lot of pages that require a login and are not the homepage. With a more accurate list of URLs, we'd be able to better assess the usage and potential breakage. Cheers :) Yoav On Wed, Jul 13, 2022 at 3:52 AM Javier Garcia Visiedo <[email protected]> wrote: Hi, I am starting a UKM collection review to opt in the existing use counter <https://chromestatus.com/metrics/feature/timeline/popularity/3873> into an UKM. Sorry if my question is too naive, I just wanted to understand what benefits would the UKM add over the existing use counter? Or is it the intention to add additional metrics to it? Thanks! Cheers Javier On Thursday, January 27, 2022 at 1:33:19 AM UTC+9 Chris Harrelson wrote: Hi, just checking in on this intent. From the API owners' perspective, we're going to wait for the UKM, thanks. On Wed, Jan 12, 2022 at 7:41 AM Yutaka Hirano <[email protected]> wrote: Hi Yoav, Thank you for the suggestions. I'll try to add UKM. > One other question that came up: Is the usage related to developers adding the "Authorization" header on their own, or is it something the browser sends under certain circumstances? (e.g. when receiving 401 responses with "WWW-Autenticate" headers) This is only for authorization headers set by scripts (via fetch() and XHR). Authorization headers the browser attaches are out of scope. On Thu, Jan 6, 2022 at 2:15 AM Yoav Weiss <[email protected]> wrote: Hey Yutaka! We discussed this at the API owners meeting today (Daniel, Chris, Alex, MikeT and myself). It seems like the risk here is too high to remove support as is, and a reasonable next step may be to add the metric to UKM and get a more detailed view of which sites are using it and how. That would enable us to better assess breakage, and reach out to those sites to reduce current usage until potential breakage reaches acceptable levels. One other question that came up: Is the usage related to developers adding the "Authorization" header on their own, or is it something the browser sends under certain circumstances? (e.g. when receiving 401 responses with "WWW-Autenticate" headers) Would renaming the headers used in such authentication protocols be a useful alternative to consider? I'm guessing that doing that would require server changes as well, so won't necessarily help with cases of existing content, but may help to move newer auth flows to make safer CORS choices. Cheers :) Yoav On Thursday, December 9, 2021 at 12:17:40 PM UTC+1 Yutaka Hirano wrote: We've been showing a deprecation message since 94 <https://chromiumdash.appspot.com/commit/9a817d69a132822ddf2954120e96a3efa2290071>. Sadly the deprecation message hasn't decreased the usage so far. On Thu, Dec 9, 2021 at 1:52 AM Mike West <[email protected]> wrote: >From my perspective, it's a bit worrying that you found user-visible breakage in a random sampling of the otherwise small number of sites that fall into this category. As Yoav suggested, there's some additional likelihood that we're not seeing some breakage that requires sign-in. It might be worthwhile to raise a deprecation warning for this behavior, and remove it after giving developers some time to adjust, perhaps with an enterprise policy for good measure. I'd be happy with a 3-release timeline, with removal thereafter. That might drive the usage down to the point where we can reasonably remove it. If it doesn't, we might need to do some more research (wiring this counter up to UKM, for instance) to see if we can track down more clear sources of potential breakage. -mike On Thursday, December 2, 2021 at 6:56:40 AM UTC+1 Yutaka Hirano wrote: On Thu, Dec 2, 2021 at 12:29 AM Yoav Weiss <[email protected]> wrote: On Wed, Dec 1, 2021 at 4:00 PM Yutaka Hirano <[email protected]> wrote: Sorry for the delay! I checked 10 sites. I saw console errors in three sites among them: 1. https://cchatty.com/ 2. https://techrxiv.org/ 3. https://bodyshake.com/ I only see a visible breakage in 1 (cards in the main panel are invisible). On other sites I don't see any visible differences. Please note that this feature is related to authorization so it is likely to break things when signing in. So it's possible that the breakage only occurs for logged in users, and is not something you'd be able to see when spot checking their homepage? Yeah, I think so. On Wed, Dec 1, 2021 at 8:11 PM Yoav Weiss <[email protected]> wrote: Friendly ping on Chris' question On Thursday, November 4, 2021 at 8:31:36 PM UTC+1 Chris Harrelson wrote: Would it be feasible to get a random list of 10-20 sites that hit the use counter and see if they are broken badly by this feature? On Thu, Nov 4, 2021 at 4:54 AM Yutaka Hirano <[email protected]> wrote: (friendly ping) On Mon, Nov 1, 2021 at 1:57 PM Yutaka Hirano <[email protected]> wrote: Thank you for the feedback. Do you have concrete steps for the investigation in your mind? On Fri, Oct 29, 2021 at 4:30 AM Mike West <[email protected]> wrote: I think it's reasonable for us to dig into the data a little bit to determine whether the 0.04% number quoted above will result in user-facing breakage. Yutaka, is that something you'd be willing to dig into? The direction seems philosophically correct to me, so I'd like to see it ship, but I'd also like to make sure we're not making the web worse for users by doing so. -mike On Thu, Oct 21, 2021 at 12:04 PM Yutaka Hirano <[email protected]> wrote: On Thu, Oct 21, 2021 at 6:25 PM Yoav Weiss <[email protected]> wrote: On Thu, Oct 21, 2021 at 9:55 AM Yutaka Hirano <[email protected]> wrote: (The implementation CL <https://chromium-review.googlesource.com/c/chromium/src/+/3226283> is under review. This intent is written as if it's landed.) Contact [email protected] Specificationhttps://fetch.spec.whatwg.org/#cors-non-wildcard-request- header-name Summary A CORS non-wildcard request header[1] is an HTTP request header which is not covered by the wildcard symbol ("*") in the access-control-allow-headers header. "authorization" is the only member of CORS non-wildcard request-header. Currently we treat the header as a usual header, which is problematic for security reasons. Implement it, and change the current behavior. 1: https://fetch.spec.whatwg.org/ #cors-non-wildcard-request-header-name Blink componentBlink>SecurityFeature>CORS <https://bugs.chromium.org/p/chromium/issues/list?q=component:Blink%3ESecurityFeature%3ECORS> TAG reviewNot needed because this implements an existing feature. TAG review statusNot applicable Risks Interoperability and Compatibility Interoperability risk is low because Mozilla and Apple showed an intent to implement this behavior. There is some compatibility risk, as the use counter[2] shows 0.04% websites would be affected. To mitigate the risk, we've shown a deprecation message for a few milestones. Can you similarly send deprecation reports as well? How long have the deprecation messages been in place? Did we see any decline in the numbers? We've shown the deprecation message since Chrome 94 <https://chromiumdash.appspot.com/commit/9a817d69a132822ddf2954120e96a3efa2290071> whose beta promotion was on Aug 26 and stable release was on Sep 21. We use CountDeprecation which sends deprecation reports automatically IIUC. I don't see any decline. Have we looked into which URLs are triggering this? (and if it's a few medium-sized properties or many tiny ones) I haven't looked at the data. Did we try outreach? No. We have an enterprise policy so that administrators can keep the existing behavior. We're planning to remove the policy on Chrome 103. 2: https://www.chromestatus.com/metrics/feature/popularity#AuthorizationCovered ByWildcard Gecko: Positive Firefox showed a positive signal in a private thread. WebKit: Positive Apple showed a positive signal in a private thread. Web developers: No signals Debuggability We'll show a CORS error to the devtools console. Is this feature fully tested by web-platform-tests <https://chromium.googlesource.com/chromium/src/+/master/docs/testing/web_platform_tests.md> ?Yes Flag nameCorsNonWildcardRequestHeadersSupport Requires code in //chrome?False (or, True only for the enterprise policy.) Tracking bughttps://crbug.com/1176753 Estimated milestones 97 Link to entry on the Chrome Platform Statushttps://chromestatus.com/ feature/5742041264816128 -- 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 on the web visit https://groups.google.com/a/ chromium.org/d/msgid/blink-dev/CABihn6G2mzUAH_Ghrqmb1xM7XetfKgB% 3DMUkX0DED7yWbL4JfGg%40mail.gmail.com <https://groups.google.com/a/chromium.org/d/msgid/blink-dev/CABihn6G2mzUAH_Ghrqmb1xM7XetfKgB%3DMUkX0DED7yWbL4JfGg%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 [email protected]. To view this discussion on the web visit https://groups.google.com/a/ chromium.org/d/msgid/blink-dev/CABihn6GB-CbbCqx0VHa7%3DLoZr%2BZN- O7o7XwTGkxXEYQa9OfupQ%40mail.gmail.com <https://groups.google.com/a/chromium.org/d/msgid/blink-dev/CABihn6GB-CbbCqx0VHa7%3DLoZr%2BZN-O7o7XwTGkxXEYQa9OfupQ%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 [email protected]. To view this discussion on the web visit https://groups.google.com/a/ chromium.org/d/msgid/blink-dev/CABihn6EqkidJF3mjfnztmMnKrEpb% 3DPTvW-kpW0s_bWCM%3DOXY8A%40mail.gmail.com <https://groups.google.com/a/chromium.org/d/msgid/blink-dev/CABihn6EqkidJF3mjfnztmMnKrEpb%3DPTvW-kpW0s_bWCM%3DOXY8A%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 [email protected]. To view this discussion on the web visit https://groups.google.com/a/chromium.org/d/msgid/blink-dev/ec699b98-abbb-4efc-8ccb-fb56098f3b29n%40chromium.org.
