[python-win32] Suggestions for a unicode build

[Roger Upole]

Mark suggested that I get some feedback from the community regarding
some of the issues involved in creating a Unicode build of win32api.

Some functions already allow you to call the ansi or wide-character version,
such as FormatMessage/FormatMessageW.  This leaves some ambiguity
as to how they should behave in a Unicode build.
Some options:
    Have only FormatMessage, calling FormatMessageW
    Leave both FormatMessage and FormatMessageW, both calling
        FormatMessagW (this would probably be the most backward compatible)
    Have FormatMessage as unicode and also add FormatMessageA in case
        there's a need to explicitely call the ansi function.

Does anyone have a preference among these, or have an alternate way to go
about it ?


It may also be possible to release the module in both wide-character and
ansi.
The question here:  Is it worth the trouble ?
The ansi version would only be needed on win98.  Python itself is dropping
support for win98 as of Py3k, but the current versions of Python are sure to
be with us for years to come.
Also, if they were released in parallel, how should they be named ?
    win32api (unicode) and win32apiA (ansi), or the other way around
    win32api (ansi) and win32apiW (unicode).


Another issue is whether or not to remove obsolete functions, specifically
the functions that deal with .ini files.  (WriteProfileSection,
WriteProfileVal, etc)
Technically, they've been obsolete since win95, but that doesn't mean noone
uses them.
The easiest thing to do for now might be to leave them as ansi, calling the
*A functions in order to minimize coding effort, and eliminate them
altogether when building for Python 3.x.


Anyone have any opinions or suggestions ?

             Roger