tools/metrics/histograms/histograms.xml

Index: tools/metrics/histograms/histograms.xml
diff --git a/tools/metrics/histograms/histograms.xml b/tools/metrics/histograms/histograms.xml
index d1f16e78ab6346bd982e90214b51f836742054e1..c2fb6a10cf2e022553ced22505a66186a2b7e6ec 100644
--- a/tools/metrics/histograms/histograms.xml
+++ b/tools/metrics/histograms/histograms.xml
@@ -6951,6 +6951,17 @@ Therefore, the affected-histogram name has to have at least one dot in it.
   </summary>
 </histogram>
 
+<histogram name="Extensions.ActiveScriptController.AlwaysRunCount"
+    units="Extension Count">
+  <owner>kalman@chromium.org</owner>
+  <owner>rdevlin.cronin@chromium.org</owner>
+  <summary>
+    The number of times the user has given a host persistent permission to run

[Ilya Sherman, 2014/08/12 00:02:11]

Hmm, the number of times since when?  The beginning of time?

It might be more appropriate to measure this user interaction by logging an
enumerated histogram that has buckets for (1) script injection request occurred,
(2) the context menu of the request was clicked, and (3) "Always Run" was chosen
from the context menu -- or something like that.  Though, whether that's better
depends on what exactly you're trying to understand about this feature.

[not at google - send to devlin, 2014/08/12 19:49:26]

That metric you mention is interesting, but I would imagine hard to analyse -
you would expect (3) to be low compared to the others given that "always run"
will only be checked once per domain, which persists, so they'd never be shown
anything again. Hm actually now that we have this feature it probably *is* worth
recording that, since you'd expect all these numbers to potentially drop, but
who knows what the correlation is due to.

Short story: I agree that we should do this, but we should log that in LogUMA
(also requires tracking it as a separate counter). I'll comment on the CL the
changes we should make for that.

*However* I'm also interested in the behaviour of users with respect to how many
they always-allow. I think that's relevant. The distribution of that would be
interesting - do users never do it (never see it, never care?). Only do it on
very few sites? Do it on most sites they visit? I'm not totally confident that
the data we get will actually be conclusive, but I'm willing to try.

[Ilya Sherman, 2014/08/12 20:00:29]

Keep in mind that if you're interested in knowing how many sites users have
whitelisted, you might want to log that based on a different trigger -- e.g.
once per UMA upload, or once per Chrome launch, or something like that.  If you
only log at the time that a user adds a site, you'll get more samples uploaded
from users who add sites more frequently.  You could certainly try to account
for that bias when viewing the data, but it might be better to try to reduce the
bias to begin with.

+    script injections. Recorded when "Always Run" is clicked in the
+    context menu of a script injection request.

[gpdavis, 2014/08/11 21:42:00]

Hopefully this is a bit more clear.

+  </summary>
+</histogram>
+
 <histogram name="Extensions.ActiveScriptController.DeniedExtensions"
     units="Extension Count">
   <owner>kalman@chromium.org</owner>