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 e70d96f68604095edb1dddee7eb53ea855694a83..8512ca8f5f839de88783249e4ea07247cfce87b2 100644
--- a/tools/metrics/histograms/histograms.xml
+++ b/tools/metrics/histograms/histograms.xml
@@ -44386,6 +44386,15 @@ http://cs/file:chrome/histograms.xml - but prefer this file for new entries.
   </summary>
 </histogram>
 
+<histogram name="NQE.RenderThreadNotified" units="count">

[Ilya Sherman, 2017/05/09 21:51:25]

nit: enum="BooleanHit" would be appropriate here.

[tbansal1, 2017/05/09 23:49:31]

Done.

+  <owner>tbansal@chromium.org</owner>
+  <owner>bengr@chromium.org</owner>
+  <summary>
+    Count of the number of times renderers were notified of the changes in the
+    network quality estimator.

[Ilya Sherman, 2017/05/09 21:51:25]

How do you plan to interpret this metric?  Is there some sort of baseline that
you are planning to compare to?  If so, could you build it into this baseline? 
Next-best option: Could you document the baseline that's expected to be used
here?

[tbansal1, 2017/05/09 23:49:31]

I plan to evaluate how frequently the notification happens over some other
metric (e.g., number of page loads or number of renderers). There is no baseline
except that we know that if the ratio is too high (e.g., 100 notifications per
renderer) or too low (< 1) then it is bad.

[Ilya Sherman, 2017/05/09 23:55:52]

Okay.  Could you please add this hint to the histogram description?  Something
like "This metric makes sense to compare in frequency relative to other events
like the number of page loads or the number of renderers seen.  It's expected
that we'll see at least 1 notification per renderer (on average), and < 100." 
(This is actually what I meant by a baseline -- the metric relative to which to
compare.  Not the expected value at the baseline =].)

Even better would be to build these events into an enum that's recorded as part
of this same metric, so that it's easy to visually compare the numbers.  But if
that requires jumping through too many hoops, then just providing a textual
reference is okay as a fallback.

[tbansal1, 2017/05/10 16:46:42]

I have expanded on the histogram description. Thanks.
I think visually comparing is okay for now.

+  </summary>
+</histogram>
+
 <histogram name="NQE.RTT" units="ms">
   <owner>bengr@chromium.org</owner>
   <owner>tbansal@chromium.org</owner>