Thanks for filing this bug!

For Media.VideoVisibleWidth, since we have Media.VideoHeight.* now, I suggest we simply drop the old one. 

+dalecurtis: WDYT?

sgtm.

For the original bug, I suggest we update the UMA to the following wherever it makes sense:
Media.Foo.MSE
Media.Foo.SRC
Media.Foo.EME
Media.Foo.ALL

This can be easily done by using the histogram suffix MediaVideoCategories [1]

Or, if it only makes sense for FFmpegDemuxer, change the name to Media.Foo.SRC

[1] https://cs.chromium.org/search/?q=MediaVideoCategories+package:%5Echromium$&ssfr=1&type=cs

 Issue 717199  has been merged into this issue.

@#4, https://codereview.chromium.org/2846693002 is taking the direction of:
Media.MSE.Foo
Media.SRC.Foo

implying that 
Media.EME.Foo and Media.ALL.Foo will have similar structure.

While the Media.ALL.Foo will obviously count instances that are also counted in MSE, SRC, or EME, will Media.EME.Foo similarly count instances that are also counted in MSE or SRC?

About where to put MSE/SRC/EME/ALL:

We actually have Media.WatchTime.MSE and Media.VideoHeight.*.MSE where MSE/SRC/ALL/EME are at the end of the UMA name. But we also have a lot of Media.MSE.* and Media.EME.* UMAs as well. So I don't have a strong preference here. Thoughts?

About meaning of MSE/SRC/EME/ALL:

ALL: everything
MSE: MSE, clear and encrypted
SRC: SRC, clear and encrypted
EME: encrypted, MSE and SRC

Likewise, I don't have a strong preference, though it seems the sequence of the bucket name is from general -> specific, favoring Media.{ALL,MSE,SRC,EME}.{<metric name like videocodec>}...

w.r.t. EME, will it ever be interesting to tease out EME-SRC vs EME-MSE for various metrics like videocodec, etc? If so, we might want to differentiate EME more precisely.

> Likewise, I don't have a strong preference, though it seems the sequence of the bucket name is from general -> specific, favoring Media.{ALL,MSE,SRC,EME}.{<metric name like videocodec>}...

I agree. The only advantage to put {ALL,MSE,SRC,EME} at the end is that we can use existing suffix so that we can avoid a lot of duplicate code in histograms.xml.

> w.r.t. EME, will it ever be interesting to tease out EME-SRC vs EME-MSE for various metrics like videocodec, etc? If so, we might want to differentiate EME more precisely.

EME is almost exclusively used with MSE. So we probably won't have too much data on EME+SRC anyways.

Assign to Dale to triage and help find a better owner on this. Thanks!

Not really anything we can do about this currently besides update the descriptions maybe. We can't rename histograms. Going forward all future ones should be clarified (and we have been doing so).

WontFix per c#13.