> So i should always specify the encoding then?
It actually depends on what further purpose you use the exported "plain text" file for. Not all plain text editors can read UTF-8 (TextWrangler, my editor of choice on the Mac can, but TextPad, the nearest Windows equivalent, cannot).
Some text editors *expect* a UTF-8 BOM code at the start of the file; others do not.
See Kasyan's struggle with uncommon characters systems in this thread: http://forums.adobe.com/thread/1141456?tstart=60
("Uncommon" only from the narrow point if view from both Microsoft and Apple -- there is nothing uncommon about Uwe, having German accented characters in his paragraph style names, and Kasyan's Cyrillic characters in same.)
Apart from always using UTF-8 (either with or without BOM), you don't have many *realistic* options. Before writing your text file, you could scan your gathered text string for any character *not* in MacRoman Encoding (this is just a list of 256 Unicodes) and replace these with a question mark. But 1. Kasyan's Cyrillic text, for example, would then be replaced in its entirety with ????'s; and, 2. you cannot properly read a MacRoman encoded text file on a Windows machine.
The same argument goes against converting to a Windows Encoding ("a" -- there are too many of them to list here).
Actually, the safest way would be to throw away everything not 7-bit ASCII. But then you'd still have an issue with line endings! Mac, Unix, Windows ...
You could try to stuff all of the above options into a dialog box and let the user decide :-/