use a locale-independent comparison

[notslang]

It's a little faster, and is more stable (since it doesn't rely upon the configured locale).

This does result in a slightly different sort order, which seems more correct. For example, with the current code, this is a sorted list:

• uber
• vice_gr
• vice_serbia
• vice_spain
• vice
• vicechina

With the new code, this is sorted:

• uber
• vice
• vice_gr
• vice_serbia
• vice_spain
• vicechina

[jerone]

Besides speed (arguable, until proven with prefs), what are your arguments to remove the locale comparison?

[notslang]

Here's a perf: http://jsperf.com/operator-vs-localecompage/14

And the other arguments are:

• It's more stable (since it doesn't rely upon the configured locale).
• It results in a slightly different sort order, which seems more correct (see first comment).

[notslang]

Is there anything else I can do to help this get merged?

[izuzak]

Is there anything else I can do to help this get merged?

@slang800 Here's a few suggestions:

1. It seems that this change doesn't break any tests -- that's surprising. In order to prevent future changes which would break the new behavior (if this PR is merged), it would be great to have tests which verify that the sort results is expected with respect to these changes.
2. A nice thing about version control is that it allows you to go back in time and see who wrote the code you're changing. That allows you to ask that person why they wrote the code like they did and ask if they have any thoughts about the change you're proposing. It would be great if you could do that here and then ping that user and perhaps also the user who made the most changes to the part of code you're changing.

What do you think?

[notslang]

rebased (someone else fixed the indentation in the time this PR has been open, so that's no longer in the diff)

[notslang]

rebased again