[webkit-dev] The case for disallowing alerts in unload, redux

[Sreeram Ramachandran]

We've discussed this in the past
(https://lists.webkit.org/pipermail/webkit-dev/2011-April/016431.html).
That thread ended with a call for data. Today, I bring examples and
statistics.

Briefly, from a histogram we added a while ago, I find that 2.3% of
all users encounter at least one such modal dialog every week. This is
much larger than I expected. I think it means that it's not just web
developers who encounter them while debugging their pages. To me, this
makes a very strong case for disallowing them.

Here are some examples of sites that show an alert when their pages
are unloaded:
http://ddl.vidplayer.net/
http://www.buenosprofesionales.com/
http://www.watchsaleshop.com/
http://52-teurseu.blogspot.com/
http://aaz-armor.blogspot.com/

The last two sites above are just two examples out of hundreds of such blogs.

Of course, alerts are also used for debugging. The proposed patch in
https://bugs.webkit.org/show_bug.cgi?id=56397 turns alerts into
console messages. Console messages can now be made to persist across
page loads, so I think this helps address the debugging use case.