On Mon, Feb 16, 2004 at 10:20:30AM -0000, Simon Marlow wrote:

...

Wolfgang Jeltsch <wolfgang@jeltsch.net> writes:

...

I meant non-ASCII characters in source code comments like this: {-| The execution time of this function is /n³/. -} Currently, Haddock seems to copy the bytes making up the non-ASCII character verbatim to the HTML file. But since the HTML file doesn't contain a character set specification, it is illformed and it depends on the browser how this situation is handled.

It shouldn't be too hard to fix this, at least for Latin-1 (full Unicode would be somewhat harder). I'll add it to the TODO list.

While Haskell's source charset is specified as Unicode, Haskell source files don't specify the byte encoding they use, so any source file using non-ASCII characters isn't portable. Entrenching Latin-1 would make the move to Unicode more difficult.

True, but GHC currently assumes Latin-1 as the encoding for source files. I don't see it as entrenching Latin-1, just that we only accept Latin-1 encoded source files; at some point in the future we might accept other encodings. Making the same simplifying assumption in Haddock doesn't seem that big a deal. Cheers, Simon