Record data used by favicons |
|||||||||
Issue descriptionWe should measure how much network data is used to fetch favicons. Favicons are typically small, e.g., 16x16 pixels, but with apple-touch-icons, they can be much larger. In addition the 404s that servers return for favicons that do not exist are often quite large themselves. UMA ideally would record favicon over-the-network sizes (it might be good to report cached favicons as having zero over-the-network size instead of ignoring them). Favicon 404s (or maybe any non-200 response) size should be reported separately. Finally, a favicon is a complete waste of data if it is never shown. It would be great to track data use for favicons that are never shown. I'm not sure what the approach would be.
,
Mar 20 2017
,
Mar 21 2017
CL in review for measuring data downloaded by favicon and its non-200 responses. https://codereview.chromium.org/2764893003/
,
Mar 21 2017
In Android, favicon is shown in the following cases. * Tab switcher * Bookmarks * History * New tab page (if the site is one of user's top visited page) * Home screen (if the user chooses 'Add to home screen') Metric should be added to track whether favicon is displayed to the user. This will determine how much of favicon data use is wasted.
,
Apr 6 2017
Hi! May I ask about the status of this effort? I see a patch with LGTMs, possibly forgotten. Thx!
,
Apr 7 2017
The following revision refers to this bug: https://chromium.googlesource.com/chromium/src.git/+/6ec568654e74b51c244d67ebb9f326fae24ec28f commit 6ec568654e74b51c244d67ebb9f326fae24ec28f Author: rajendrant <rajendrant@chromium.org> Date: Fri Apr 07 20:05:24 2017 Record data use of favicon This CL records the network data use for downloading favicons. Also the size of non 200 responses for favicon requests are recorded. BUG= 698347 Review-Url: https://codereview.chromium.org/2764893003 Cr-Commit-Position: refs/heads/master@{#462975} [modify] https://crrev.com/6ec568654e74b51c244d67ebb9f326fae24ec28f/components/data_use_measurement/content/content_url_request_classifier.cc [modify] https://crrev.com/6ec568654e74b51c244d67ebb9f326fae24ec28f/components/data_use_measurement/content/content_url_request_classifier.h [modify] https://crrev.com/6ec568654e74b51c244d67ebb9f326fae24ec28f/components/data_use_measurement/core/data_use_measurement.cc [modify] https://crrev.com/6ec568654e74b51c244d67ebb9f326fae24ec28f/components/data_use_measurement/core/data_use_measurement_unittest.cc [modify] https://crrev.com/6ec568654e74b51c244d67ebb9f326fae24ec28f/components/data_use_measurement/core/data_use_network_delegate_unittest.cc [modify] https://crrev.com/6ec568654e74b51c244d67ebb9f326fae24ec28f/components/data_use_measurement/core/url_request_classifier.h [modify] https://crrev.com/6ec568654e74b51c244d67ebb9f326fae24ec28f/tools/metrics/histograms/histograms.xml
,
Apr 7 2017
From mastiz@, pkotwicz@: * NewTabPage.SuggestionsImpression counts tile impressions in the NTP * UserActions.Counts.MobileToolbarShowStackView gets recorded each time that the tab switcher button is pressed I will start analyzing these metrics, once some data obtained from commit c#6.
,
Apr 30 2017
We should have some data now. Please analyze.
,
May 3 2017
Based on Beta channel and metric DataUse.FavIcon.Downstream, what I see is roughly: - 50th percentile: ~1 KB. - 95th percentile: 13 KB. - 99th percentile: 40 KB. The other takeaway is that the zero bucket (representing HTTP cache hits IIUC) represents 30% of all requests.
,
May 3 2017
I am a bit puzzled: is it size per one icon? Can we somehow infer daily/weekly/monthly traffic for one user? (if not, wouldn't be of interest to record such things for all the DataUse types, bengr@?)
,
May 3 2017
Do you mean percentiles for sums per user? Averages are obviously there since we know the number of users. Anyway, total sums is likely what we need. As per the original plan, I believe we were going to compare these metrics (sums) with some other number of interactions (sums), as per c#7.
,
May 3 2017
Yes, I meant recording the traffic per user in order to see percentiles. I agree this is not needed for this particular bug. In general, this would be nice. I am just curious (mwd@, bengr@) - can I obtain the percentiles for sums per user using collab from the existing metrics or - would separate metrics be needed for that?
,
Nov 8 2017
Raj, what's the status?
,
Dec 27 2017
,
Feb 13 2018
Raj, anything left to do on this one? Or is it done?
,
Feb 13 2018
No code changes necessary. Some in-depth analysis is pending and I could close it with that update.
,
Apr 11 2018
Any updates? Can we just close this?
,
Apr 11 2018
,
Sep 21
We need to do some analysis to verify that it's logged properly.
,
Sep 21
,
Jan 18
(5 days ago)
Usage of favicons is pretty low. https://uma.googleplex.com/p/chrome/histograms?endDate=20190116&dayCount=7&histograms=DataUse.FavIcon.Downstream%2CDataUse.FavIcon.Downstream.Non200Response&fixupData=true&uniqueUsers=true&showMax=true&filters=platform%2Ceq%2CA%2Cchannel%2Ceq%2C4%2Cisofficial%2Ceq%2CTrue&implicitFilters=isofficial |
|||||||||
►
Sign in to add a comment |
|||||||||
Comment 1 by mastiz@chromium.org
, Mar 6 2017