Re: patch -- fix resolution in pango (win32 backend)

[Owen Taylor <otaylor redhat com>]

OK, this thread has gotten far too long :-). I think the choice is
straightforward, if not simple:

 a) 12 points means the same thing as it does in simple 
    Windows apps.

 b) 12 points means the same thing as it does for Pango on 
    Unix.

Whether the Windows or Pango-on-Unix setup is better setup is
something I don't have time to get into at the moment, and not
particularly relevant. (I will say its a shame that people who want
bigger fonts on Windows frequently decrease their resolution to get
them...)

Consider various applications:

 - Fonts built into an app. Apps shouldn't be doing this; fonts
   should be based on the system fonts size, so it doesn't 
   matter. 
 
   (gdk/win32 can do the logical => physical transformation
   in the code to default-system-font preference.)

 - Fonts specified by the user. It's probably good if picking
   a "12 point" font gives the same result in a Gtk/Win32 app
   as in another Win32 app.

 - Fonts in a "document" setting. If the app wants to control
   the points => pixel transformation, then it needs a predictable
   expectation of how it will work, independent of platform.

   The only way we expose the resolution in GDK is 
   gdk_screen_width_mm(), so Pango needs to be consistent with
   this. We can do to this; either make Pango use the physical
   resolution, or make gdk_screen_width_mm() use the logical
   resolution.

I think on balance, we should:

 * Use the logical resolution in Pango
 * Make the the results of gdk_screen_width_mm() reflect the logical
   resolution rather than the physical resolution.

Regards,
                                        Owen