[Haskell-cafe] Problem with text and text-icu packages

Hi, I'm struggling with text-0.1 and text-icu-0.1, as announced at http://www.serpentine.com/blog/2009/02/27/finally-fast-unicode-support-for-h... The code in http://pastebin.com/m7d8d9f91 is intended to read in a UTF-8 file a1.txt, reverse it twice, and write it out to another UTF-8 file a2.txt, which I would have thought should make a2.txt identical to a1.txt. Mostly it is, but sometimes it isn't, e.g.: % echo a > a1.txt % tryicu # <- my code % cat a2.txt ࠋ퐤 Specifically, a1.txt contains bytes 97 and 10, while a2.txt contains bytes 224 160 139 237 144 164. Have I misconstrued things, or is this a bug? I am running ghc 6.10.2 and ICU 4.0.1 on SuSE. Thanks for any help David