Hello Streams library includes CharEncoding transformer that allows to read/write files in UTF-8 and any other encodings. according to the Einar's requirement, i made CharEncoding transformer dynamic - apllied encoding can be changed at any time. now i thinks that this design may limit speed of encoding and plan to use alternate design where concrete encoding is "burned" into the tramsformer and cannot be changed after creation of the transformed stream. due to the ability to create transformed stream at any moment this should not impose big limitations. say, instead: h <- openFD "test" WriteMode >>= withEncoding utf8 vPutStrLn h "testing" vSetEncoding h latin1 vPutStrLn h "testing" now we should write: h <- openFD "test" WriteMode h1 <- withEncoding utf8 h vPutStrLn h1 "testing" h2 <- withEncoding latin1 h vPutStrLn h2 "testing" or h <- openFD "test" WriteMode blockWithEncoding utf8 h $ \h -> do vPutStrLn h "testing" blockWithEncoding latin1 h $ \h -> do vPutStrLn h "testing" So, the gain is faster speed on typical usage where dynamic change of encoding is not used. the loss is more restricted abilities to change the encoding - the stream with new encoding is new object and therefore these encoding changes should have a structure of enclosed blocks (such as HTML structure) THE QUESTION: is this new interface enough for your usage, or someone will need a more dynamic approach given by the old interface? -- Best regards, Bulat mailto:bulatz@HotPOP.com