"John.Cowan" <xxxxxx@reutershealth.com> writes:

> Thomas Bushnell BSG scripsit:
>
>> A generically named function on a fancy/schmancy
>> system should do the correct locale-dependent operation when case
>> mapping is requested.  It should also provide a "neutral" locale which
>> will implement the locale-independent case mapping from the Unicode
>> data files.
>
> Then this seems to be a matter of taste (or theology): you think that
> the names which our fathers knew of old ought to be bound to the smartest
> routines available in a given implementation, whereas I think they ought
> to be bound to simple basic universally definesd routine that does all
> of the job in some circumstances and part of it in others.  If you agree
> that this is the remaining point of disagreement, I'll say no more about
> it, and those whose job it is to decide can do so.

I agree that this is so, but I would not use the same words to
describe it. :)

I'll say *why* I think this way though:

Programmers are very likely to use the standard functions when they
can; they are very likely to do so.  I want them to automagically get
the fancy/schmancy things, which seems the only right behavior on a
system with fully integrated multi-language support.  They will not
*want* to write something with half-assed support, but they are very
likely to be misled.

In other words, I want the "names which our fathers knew of old" to be
bound to something which is not idiosyncratic, or represents only a
subset of the system.

>> As long as the functions do not rigidly require
>> specific behaviors that are known to cause problems, and are not
>> misleadingly named, I'm content.
>
> All specific behaviors are known to cause problems in *some* circumstances.

"and are not misleadingly named" which is not a separate condition,
but one working together with the first.

If you don't name it misleadingly, then I don't care what the function
does here.