some patches and ideas

[shula]

ah, shit -- here's the attachment

addition to #1:
the mis-encoded word seems to be אבד, which is in this case the grammatical-root of the word.
it seems that somewhere on the way (either babylon or GD) decode this as LATIN-1.

[dmdmdm]

my patch reverses the "root" string and adds brackets around it. then it is added to the definition which is encoded later (after all of the data of this definition is read) according to its charset - so in English definition, it is encoded (incorrectly) with English charset.

[masterman]

I think the efforts to making RTL support for Hebrew language can be applied to other RTL languages like Arabic, Farsi and Kurdish!
If it occurs, It would be amazing!
I think it's not difficult to do the same works for Arabic, Farsi & Kurdish like Hebrew. So I offer someone to do it.
My C++ knowledge is in good condition (I only think so), but I never contributed in a big project, specially an open source one. I may work on it myself but only a try, maybe successful!