Re: longtasks & hidden windows

On Tue, May 2, 2023 at 5:44 PM Jansma, Nic <njansma@akamai.com> wrote:

> Hi Noam!
>
>
>
> In the boomerang.js library (and thus in mPulse RUM data), we use
> LongTasks to generate an aggregate RUM *Long Task Time* (LTT) metric and
> as an input to our *Time To Interactive* (TTI) metric.
>
>
>
> Looking at our data for today, about 13% of the page loads we see are
> marked as fully "hidden", i.e. for as long as boomerang.js is loaded, the
> page is backgrounded or minimized.  It sounds like in those cases, we would
> not get any LongTasks reported, and our TTI metric may not work (or would
> report better-than-expected results).  This could cause some shifts in
> reported data once this version of Chrome is shipped, so we'd likely want
> to document the change for our customers.
>
Understood. We are still experimenting with this.


>
>
> One other thought – if LongTasks reporting would start/stop as a page
> transitioned between visible and not (switching between tabs during page
> load, for example), might it confuse analytics in other ways?  For example
> if RUM sees LongTasks for some time, then none, then sees LongTasks again,
> etc – the periods of "hidden" may be reported as "smooth-sailing" where it
> just wasn't generating any LongTasks data (where it would today).
>
> You'll see the "visibility-state" entry which would allow you to discern
that that's the reason for seeing no long tasks!


>
>
> We can adapt either way, i.e. we may consider disabling TTI calculation if
> the page is transitioning multiple times between hidden/invisible.
>
Thanks for the input :)


>
>
> Nic Jansma
>
> [image: cidimage001.png@01D68B7B.AE0304C0]
>
> he/him
> +1.408.650.0848 x70848 <(408)%20650-0848>
>
> Akamai Technologies
> 150 Broadway
> Cambridge, MA 02142
>
>
>
>
>
> *From: *Noam Rosenthal <nrosenthal@google.com>
> *Date: *Monday, May 1, 2023 at 10:17 AM
> *To: *"public-web-perf@w3.org" <public-web-perf@w3.org>
> *Subject: *PSA: longtasks & hidden windows
> *Resent-From: *<public-web-perf@w3.org>
> *Resent-Date: *Monday, May 1, 2023 at 10:16 AM
>
>
>
> Hola webperfies
>
> I'm planning to introduce a change to the implementation of longtask
> reporting in chrome, that would fix a few longtask-related issues, but
> would introduce a new change in behavior: long tasks would only be reported
> for visible windows.
>
>
>
> Bug: https://bugs.chromium.org/p/chromium/issues/detail?id=1441419
> <https://urldefense.com/v3/__https:/bugs.chromium.org/p/chromium/issues/detail?id=1441419__;!!GjvTz_vk!WWEbBtKVU30R4-R-tF9PLs5OrO2tUo0amdp91SRdCTB4IgpmqAPW0J31kNxOci4xMyQzKhYsO5nvmrdrcA$>
>
> CL (behind a flag):
> https://chromium-review.googlesource.com/c/chromium/src/+/4495305
> <https://urldefense.com/v3/__https:/chromium-review.googlesource.com/c/chromium/src/*/4495305__;Kw!!GjvTz_vk!WWEbBtKVU30R4-R-tF9PLs5OrO2tUo0amdp91SRdCTB4IgpmqAPW0J31kNxOci4xMyQzKhYsO5kF7XrwGw$>
>
>
>
> Before moving forward with this, I'd like to surface this to the group and
> see if anyone is relying on longtasks being reported to hidden windows
> (e.g. minimized, background tab).
>
>
>
> Thoughts? Objections? 👍s?
>