Mlterm & Arabic

[<mig at entrisphere dot com>]

Salaam Aleikum All,
	I am attempting to teach myself arabic, and have created a bunch
of programs aimed at testing my vocabulary. Up until now, I have been
running these programs under acon and all was well. I have recently got
mlterm running under Linux (RH8). I have seen two problems that I am
having difficulty tracking down, and was wondering if anyone could help.


i) When being tested Arabic to English, the form of the line is as
follows:

	1) XXXX:

	Where XXXX is the arabic word (I type the English transaltion at
the colon) Since 1 is a weak unicode character, I need to put in a LRM
character to force left to right ordering. The problem is that mlterm
echoes a box character to the screen instead of ignoring the LRM
character. With debugging on I have the following output: (the word I am
being testing on is mahkamapN (buckwalter notation))

utf8 text =>
2000e 0031 0029 0020 0645 062d 0643 0645 0629 003a 0020 0020
fribidi_remove_bidi_marks => removed 1 chars
0031 0029 0020 0645 062d 0643 0645 0629 003a 0020 0020
Visual order => 00 01 02 03 08 07 06 05 04 09 10

As you can see, the visual order is correct (I threw in a call to
fribidi_remove_bidi_marks, and printed the output). Clearly mlterm is
trying to print the '200e' (LRM) character. Is this a bug? Has anoyone
seen this before?

ii) When being tested English to Arabic, after I flip the keyboard
mapping and begin entering Arabic letters, I can backspace over the
english. It looks like flipping the keyboard introduces an ambiguity
into the ordering, and spaces! Are being added to the input.


Any help would be great,
				Thanks,

					Matt

Matt Greenwood,
Entrisphere Inc., 
2770 San Tomas Expressway,
Santa Clara, CA, 95051
(Tel) 408-850-0343
(Fax) 408-970-8373