RE: Character folding in the pretest

[Drew Adams]

> > Е and Ё, on the other hand, are a holywar-inducing contention
> > point.
> 
> They have their own places in the Russian alphabet.  I think
> char-folding should fold only "modified" letter variants into
> "canonical" form (without any modifications).
> 
> Е and Ё are just separate letters, although we don't use Ё much...
> 
> Once I "fixed" all our text resources files at work and a colleague of
> mine commented in review that Ё is used only in childrens books.  I had
> to revert the change.

The point, IMO, is that there are multiple use cases, depending on
the user and the context (including, but not limited to, language).

What we really need are ways for _users_ to _easily_ express their
preferences, including perhaps preferences for different contexts
that they use, and including ways to express what they want on the
fly - not just ahead of time via Customize (e.g. default preferences).

That should be the _first_ order of business.  If we do a good
job of providing for that then anything additional we do
concerning DWIM or default behaviors is icing on the cake.

If we do not take care of the need to give users flexible control
then anything we do (DWIM or defaults) will be misguided for at
least some users and use cases.  It typically hurts more than helps,
IMO.

This is a general point, not limited to char folding or search.
Our priority should be to (1) yes, raise possible use cases for
discussion, such as is being done now in this thread, and (2)
come up with brilliant, easy-to-use ways to _give users control_.

Users are different, and even the same user has multiple use
cases - s?he does not want the same behavior all the time.
It is not enough to look at the user's language setting etc.
Only the user knows, at any given time, what s?he wants.

It is fine to be smart about the defaults we set, but that's
not the most important thing.  Likewise wrt coming up with
clever DWIM behavior.  But the smartest DWIM is brain dead
when compared with a live user.  And even the best default
behavior is no good for many use cases.  Users need to be
able to (easily) control the behavior.

Thinking first about defaults or DWIM is wrong, IMO.  We
should think first about how users can change the behavior,
including on the fly.