Variables and casing, again

Looks like WebKit recently checked in a patch [1] that does some sort of 
case-insensitive stuff on variables, via lowercasing them and then 
comparing lowercased versions.

It's not quite clear to me what lowercasing algorithm it's using; at 
first glance it's passing UTF-16 code units one by one to a function 
that expects to get 32-bit codepoints.  But maybe that's just me 
misreading the code.

In any case, whatever it's doing doesn't sound like anything we've 
discussed on this list, since, again if I read the code right, it'll 
treat I (U+0049) and İ (U+0130) as equal.

Can we come to a decision on this before UAs start shipping random gunk 
like that and causing backwards-compat problems?  :(

-Boris

[1] http://trac.webkit.org/changeset/131313