accelerator keys for non-ASCII character languages

[raphael cs uky edu (Raphael Finkel)]

Is there any consensus on how do establish accelerator keys for languages
(Hebrew, Yiddish, Arabic, Farsi, Hindi, Russian, Armenian, ...)  that don't use
Ascii?  I can easily use & in my translations, but the effects are poor.

1.  The underlining that indicates that a character is an accelerator conflicts
with a vowel mark in Hebrew/Yiddish, which is also a horizontal line under the
letter, as in אַ.  They look quite similar.  I don't know how to tell the
software to use a different indicator for the shorthand.  (Different color
would be a possibility, or surrounding the character in a box, or changing the
background color for that character.)

2. Some characters that I would like to use as accelerators, such as פֿ, are
formed as the composition of a base character and one or more combining
characters.  Gtk+ displays &bc (where b is base and c as combining, here פֿ) as
b with an underscore and then just c in the next position (here פ̲ ֿ), which is
wrong.  In other words, only the first character (פ) is "accelerated", without
the combiner.

3.  It appears that Gtk+ input methods are not invoked on <alt-char>
combinations, so I can't feed keystrokes through a Gtk+ input method to get
acceleration.

4.  I can use xmodmap to mostly remap my keyboard into the Yiddish character
set.  This method works.  But using xmodmap to access accelerator keys is bad
for several reasons: (a) There are Yiddish characters, such as ױ, that have no
entry in the X Windows tables, so there is no way to generate a mapping.  (b) I
prefer to use a Gtk+ input method (either mine or Noah's) to convert Ascii
keystrokes into sometimes complex Unicode sequences.  I can't use the input
methods and a remapped keyboard at the same time; they represent two different
approaches to getting non-Ascii input.  It is a pain to switch between them,
more than the pain of using the mouse to click an entry instead of using a
keyboard shortcut.

5.  One suggestion is to place an Ascii accelerator even on non-Ascii menu
entries, such as "טעקע (&F)".  The result looks ok on pop-up menus, but
terrible on toolbars.  There is no easy way for me to tell as I enter
translations how a particular message will be used.

6.  The upshot is that in my translations, I have been leaving out all keyboard
acceleration.  It's a pity.  Suggestions?

Raphael Finkel