Track per-user actives in HKLM...ClientStateMedium for system-level installs |
||||||
Issue descriptionUse the same aggregation strategy as is used for _NumAccounts and _NumSignedIn. Perhaps _DidRun=1 when setting the value, and _DidRun=0 when clearing it. For user-level installs, write _DidRun into ClientState alongside dr. Initially, make this addition without removing the old code so that we can compare the two. Eventually delete the old method. Ganesh: please let me know if there are any special considerations, or if this doesn't quite capture your request. Thanks.
,
Sep 5 2016
,
Sep 6 2016
I checked the code, and currently, App Defined Attributes are not reset a la ApplicationUsageData::ResetDidRun. I presume this would be a problem?
,
Sep 6 2016
The following revision refers to this bug: https://chromium.googlesource.com/chromium/src.git/+/f7c30a5abe40796ca0972f541324c6a117d3ca64 commit f7c30a5abe40796ca0972f541324c6a117d3ca64 Author: grt <grt@chromium.org> Date: Tue Sep 06 20:52:22 2016 Generalized the storage strategy used by UpdateProfileCounts. This simplifies adding new such stats. BUG= 643647 TEST=covered by GoogleUpdateSettingsTest.UpdateProfileCounts* Review-Url: https://codereview.chromium.org/2308423002 Cr-Commit-Position: refs/heads/master@{#416724} [modify] https://crrev.com/f7c30a5abe40796ca0972f541324c6a117d3ca64/chrome/installer/util/google_update_settings.cc
,
Jun 29 2018
felt@ + mpearson@: we discussed a while back whether we should add a new metric to count users in addition to our current strategy of counting only installs. Have either of you thought about this at all since then? Please either re-assign this back to me (if we should move forward) or mark it as WontFix (if we don't need a per-user active count). Thanks.
,
Jun 29 2018
Help me page this back in my memory. How far ago was "a while back", and what do you (did we) mean as "users": profiles, operating system users, or GAIA based on Chrome signed-in status?
,
Jul 2
,
Jul 9
Assigning back to grt@ for clarification; removing from my queue.
,
Aug 1
We talked it over at the analysis meeting this week and decided we don't feel a great need to know number Chrome users on machines with system level installs.
In brief, the justification is that both ways of counting ("installs" versus "OS-level users") come with caveats. We not convinced that switching would make any of the metrics we look at any cleaner or make them easier to understand, so we're fine with sticking with what we already have.
,
Aug 2
SGTM. Thanks for discussing and getting back to me. |
||||||
►
Sign in to add a comment |
||||||
Comment 1 by grt@chromium.org
, Sep 5 2016