906032 - Export Telemetry data CHARSET_OVERRIDE_USED by locale

Status: RESOLVED FIXED

(Mozilla Metrics :: Data/Backend Reports, defect)

Description

[Mark Reid [:mreid]]

Provide a by-locale breakdown of the telemetry flag CHARSET_OVERRIDE_USED on the release channel since Firefox 22 in a) desktop Firefox and b) Firefox for Android to inform the inclusion of the character encoding over a feature in a) Australis and b) B2G.

Data to be provided in a form like:

date, locale, app, on_count, off_count
2013-08-16, pt-BR, Firefox, 4, 500

Comment 1

[Mark Reid [:mreid]]

Attachment: bug906032.20130818.clean.txt

Here is one day's exported data.  I will export more days if this looks OK.  I added in the Firefox version - if you don't want/need that, I can take it out.

Comment 2

[Henri Sivonen (:hsivonen)]

This looks great. Thank you.

Comment 3

[Henri Sivonen (:hsivonen)]

Comment on attachment 792831 [details]
bug906032.20130818.clean.txt

Oops. I didn't realize there was another pending flag as well. Sorry.

Comment 4

[Mark Reid [:mreid]]

Attachment: Flag usage from 20130807 to 20130826

Here is an export of the data going back to August 8th.  I will export more data and attach it here as it becomes available.

Please let me know once you've seen enough :)

Comment 5

[Henri Sivonen (:hsivonen)]

Attachment: Python script for understanding the data

Usage:
cat *.txt | python count_charset.py > result.html

Comment 6

[Henri Sivonen (:hsivonen)]

Attachment: Summary data as HTML

Comment 7

[Henri Sivonen (:hsivonen)]

(In reply to Mark Reid [:mreid] from comment #4)
> Please let me know once you've seen enough :)

I think the data you attached is sufficient. Thank you!

Comment 8

[Henri Sivonen (:hsivonen)]

Comment on attachment 796536 [details]
Summary data as HTML

Some observations: Unsurprisingly, locales that don't heed to the advice given in https://developer.mozilla.org/en-US/docs/Localizations_and_character_encodings are the locales that sort to the top by character encoding overwrite usage. Next up is Traditional Chinese. My guess is that people who use Traditional Chinese browser UI read Simplified Chinese content often enough that the difference in the legacy encodings between the two explains the need to use the menu.

I guess for starters, I should file bugs on the locales that default to UTF-8 and are high up on the list and on the locales whose fall back is listed as "(broken value)".