New UseCounter data now appearing on chromestatus.com

[Rick Byers]

tl;dr: There's a 3x-7x jump in the absolute values of most of the usage stats shown on chromestatus.com.  This is expected and believed to be a more accurate representation of true usage. Since it's generally only order-of-magnitude-level comparisons that we focus on for removal decisions, we don't think will have much impact on future (or past) compat analysis.

---

Back in January I rolled out new UseCounter histograms (Blink.UseCounter.*) designed to address a number of flaws in the original ones (WebCore.FeatureObserver).  Eric and I have now updated chromestatus.com such that new data uses these new metrics. Unfortunately this means there's a sharp jump on the graphs at 10/26. Eric has added a UI afforadance to help explain this.

At the same time I've re-worked the data pipeline (internal doc) such that every data point represents a single Chrome milestone only (that which has the most usage of the given feature). So the data should now (mostly) be useful for compat purposes even when a counter hasn't yet hit stable, or has been removed from the current stable. Eric is still working on some UI  enhancements to better expose this new information. As always, Googlers may want to leverage the internal UMA dashboard for more sophisticated analysis including data we're not currently able to share publicly (breakdown of mobile vs. desktop, etc.).

Looking forward, Luna is working on migrating UseCounter collection from blink to the browser's page load metrics infrastructure (to properly support OOPIFs and enable more sophisticated types of analysis). This will involve a few other tweaks to the semantics, but we don't expect any significant on most of the numbers.

Let me know if you have any questions or concerns.

Thanks,

   Rick

[Daniel Bratell]

More accurate numbers are good!

In some cases it changes how I think about features. I see Quirks mode going from 3% to 12%, so much more common than people might have thought before.

Just randomly clicking through them, I see a lot of them are at 0 (or close enough to be indistinguishable). Would every metric at 0 be a potential feature removal?

/Daniel

--
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+...@chromium.org.
To view this discussion on the web visit https://groups.google.com/a/chromium.org/d/msgid/blink-dev/CAFUtAY-Zq5Bs76p3nfC1etvmeNkrA_HytLfc7-oECc9nLB0t_g%40mail.gmail.com.

--

/* Opera Software, Linköping, Sweden: CET (UTC+1) */