LGTM1 I'm excited about this work as it unlocks many potential wins in terms of loading performance. Interest from Mozilla as well as support from leading CDNs is also promising in terms of activation and usage. Thanks for working on this!!
On Wed, Apr 27, 2022 at 7:49 AM Daisuke Enomoto <[email protected]> wrote: > Hi > > Any other questions or concerns? > > Thanks! > > On Thu, Apr 21, 2022 at 5:23 PM Kenichi Ishibashi <[email protected]> > wrote: > >> Hi Yoav, >> >> On Thu, Apr 21, 2022 at 3:26 PM Yoav Weiss <[email protected]> >> wrote: >> >>> >>> >>> On Thu, Apr 21, 2022 at 4:13 AM Kenichi Ishibashi <[email protected]> >>> wrote: >>> >>>> Contact emails >>>> >>>> [email protected], [email protected] >>>> >>>> Explainer >>>> >>>> https://github.com/bashi/early-hints-explainer/blob/main/explainer.md >>>> >>>> Specification >>>> >>>> https://tools.ietf.org/html/rfc8297 >>>> >>>> https://html.spec.whatwg.org/#early-hints >>>> >>>> Design docs >>>> >>>> >>>> https://docs.google.com/document/d/1gCh_CnfrJq_VL7aGoq6skc7sn4yn5pKsM0gkHe5B9go/edit?usp=sharing >>>> >>>> Summary >>>> >>>> Support 103 Early Hints informational responses during top-level frame >>>> navigation. When a 103 response includes link headers Chromium tries to >>>> preload specified resources (including modulepreload) or preconnect to >>>> specified origins before the final response is received. This gives Web >>>> developers a way to optimize Core Web Vitals such as Largest Contentful >>>> Paint (LCP). >>>> >>>> >>>> Blink component >>>> >>>> Internals>Preload >>>> <https://bugs.chromium.org/p/chromium/issues/list?q=component:Internals%3EPreload> >>>> >>>> TAG review >>>> >>>> https://github.com/w3ctag/design-reviews/issues/638 >>>> >>>> TAG review status >>>> >>>> Issues addressed >>>> >>>> Risks >>>> Interoperability and Compatibility >>>> >>>> Browsers may not understand a 103 response but they are expected >>>> <https://www.ietf.org/archive/id/draft-ietf-httpbis-semantics-19.html#name-informational-1xx> >>>> to ignore 1xx responses if they don’t understand 1xx responses. Firefox is >>>> implementing 103 support. >>>> >>>> Proxies may not understand a 103 response. They may treat the 103 >>>> response as a part of the final response when the response is sent over >>>> HTTP/1.1. The problem is less likely to happen over HTTP/2 and HTTP/3 >>>> thanks to their frame format. Chromium only handles 103 responses over >>>> HTTP/2 and HTTP/3. >>>> >>>> The HTML and Fetch standards have been updated to handle preload link >>>> headers in 103 responses for navigations (whatwg/html#7675 >>>> <https://github.com/whatwg/html/pull/7675>, whatwg/fetch#1404 >>>> <https://github.com/whatwg/fetch/pull/1404>). For preconnect there are >>>> ongoing efforts to specify the behavior. See “Anticipated spec changes” >>>> section below for more information. >>>> >>>> Gecko: In development >>>> <https://bugzilla.mozilla.org/show_bug.cgi?id=1407355>. >>>> >>> >>> Can you ask for an official position? https://bit.ly/blink-signals >>> >> I should have put the link to the official position. They said it's worth >> prototyping <https://github.com/mozilla/standards-positions/issues/134> and >> IIUC it's under development. >> >> >>> >>>> >>>> WebKit: No signal >>>> <https://lists.webkit.org/pipermail/webkit-dev/2021-May/031861.html>. >>>> >>>> Web developers: Positive. Positive interest and intent of support by >>>> popular CDNs (Fastly >>>> <https://www.fastly.com/blog/faster-websites-early-priority-hints>, >>>> Cloudflare <https://blog.cloudflare.com/early-hints/>). We've been >>>> collaborating with partners to evaluate the impact of Early Hints and we >>>> observed significant improvements on page loading metrics such as FCP. >>>> >>>> >>>> Ergonomics >>>> >>>> The primary usage of this feature is to send link headers to preload >>>> resources or preconnect to servers to improve page load performance. Due to >>>> potential security concerns Chromium only handles the first 103 response >>>> for the top-frame navigation. >>>> >>> >>> That restriction is specified, right? AFAICT, "process early hints >>> headers >>> <https://html.spec.whatwg.org/multipage/semantics.html#process-early-hint-headers>" >>> is only called from "process a navigate response >>> <https://html.spec.whatwg.org/multipage/browsing-the-web.html#navigating-across-documents:process-early-hint-headers> >>> ". >>> >> Yes, you're right. Sorry for the confusion. Chromium has these >> restrictions to follow the specification. >> >> >>> >>> >>>> See this document >>>> <https://chromium.googlesource.com/chromium/src/+/master/docs/early-hints.md> >>>> for details of restrictions. These restrictions may be relaxed in the >>>> future. >>>> >>>> >>>> Activation >>>> >>>> Popular HTTP servers and CDNs provide configurations or options to send >>>> Early Hints responses. >>>> >>>> >>>> Security >>>> >>>> Request mode <https://fetch.spec.whatwg.org/#concept-request-mode> and >>>> credentials >>>> mode <https://fetch.spec.whatwg.org/#concept-request-credentials-mode> >>>> are calculated the same as normal preload and preconnect. Preloads >>>> triggered by a 103 response follow policies specified in the same 103 >>>> response. Chromium only handles the first 103 response and ignores >>>> following 103 responses to avoid contradiction policies between these >>>> responses and the final response. Chromium doesn’t use preloaded resources >>>> triggered by a 103 response when there are conflicting policies between a >>>> 103 response and the final response. See the security considerations >>>> section >>>> <https://docs.google.com/document/d/1gCh_CnfrJq_VL7aGoq6skc7sn4yn5pKsM0gkHe5B9go/edit#heading=h.fhq04xt0tah6> >>>> of the design doc for more information. >>>> >>> >>> Is that specified? >>> >> Yes. These behaviors are specified in "process a navigate fetch >> <https://html.spec.whatwg.org/multipage/browsing-the-web.html#process-a-navigate-fetch>" >> and "process early hint headers >> <https://html.spec.whatwg.org/multipage/semantics.html#process-early-hint-headers> >> ". >> >> >>> >>>> >>>> >>>> WebView application risks >>>> >>>> We don’t expect WebView specific problems with this feature. >>>> >>>> >>>> Debuggability >>>> >>>> If a resource is preloaded by Early Hints, the corresponding >>>> PerformanceResourceTiming object reports that its initiatorType is >>>> "early-hints". >>>> >>>> When the document doesn’t use resources preloaded by Early Hints within >>>> a few seconds from the window’s load event Chromium will show warning >>>> messages (this is under review now). >>>> >>>> DevTools lack Early Hints support and we consider adding Early Hints >>>> support in DevTools in the future. In the meanwhile developers can use >>>> NetLog >>>> <https://www.chromium.org/for-testers/providing-network-details/>, >>>> WebPageTest, or network protocol analyzers such as Wireshark for debugging. >>>> >>>> Is this feature fully tested by web-platform-tests >>>> <https://chromium.googlesource.com/chromium/src/+/master/docs/testing/web_platform_tests.md> >>>> ? >>>> >>>> Yes >>>> <https://github.com/web-platform-tests/wpt/tree/master/loading/early-hints> >>>> >>>> Flag name >>>> >>>> EarlyHintsPreloadForNavigation >>>> >>>> Requires code in //chrome? >>>> >>>> False >>>> >>>> Tracking bug >>>> >>>> https://crbug.com/671310 >>>> >>>> Launch bug >>>> >>>> https://crbug.com/1197989 >>>> >>>> Sample links >>>> >>>> https://early-hints.fastlylabs.com/ >>>> >>>> Estimated milestones >>>> >>>> We would like to ship this feature in M103. >>>> >>>> Anticipated spec changes >>>> >>>> Historically preconnect hasn’t been specified well (whatwg/html#4224 >>>> <https://github.com/whatwg/html/issues/4224>). There are ongoing >>>> efforts to specify the behavior. For example whatwg/html#7814 >>>> <https://github.com/whatwg/html/pull/7814> specified the behavior of >>>> <link rel=preconnect>. whatwg/html#7835 >>>> <https://github.com/whatwg/html/pull/7835> will specify preconnect in >>>> link headers and it will cover 103 responses. >>>> >>> Great to see those PRs have landed! > >>>> >>>> Link to entry on the Chrome Platform Status >>>> >>>> https://chromestatus.com/feature/5207422375297024 >>>> >>>> Links to previous Intent discussions >>>> >>>> Intent to Experiment: >>>> https://groups.google.com/u/1/a/chromium.org/g/blink-dev/c/xb_NHDS3twY >>>> >>>> >>>> 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 on the web visit >>>> https://groups.google.com/a/chromium.org/d/msgid/blink-dev/CAPLXX-9MUr5cdGw_x7Gy6B3dxcmpVp9tnXL%2BY%3DDeEXoABNy3bA%40mail.gmail.com >>>> <https://groups.google.com/a/chromium.org/d/msgid/blink-dev/CAPLXX-9MUr5cdGw_x7Gy6B3dxcmpVp9tnXL%2BY%3DDeEXoABNy3bA%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/CAL5BFfU3vO5hb3CXXt4mG6HCWPOF7tjCoqvrjOdbW2Q49uiSbw%40mail.gmail.com.
