Simon Peyton-Jones wrote:
> One difficulty is that the definition tries not to be Unix-centric.
This is exactly the point I would make. It would however be nice
if there were a separate chapter of the standard for Posix-compliant
OS's (and perhaps also one for Windows) which would document GHC's
Posix functions and also clarify issues such as this. SML does this
with its library functions. Implementations would be free to implement
any or none of these chapters. Some could implement more than one,
for example where there is a Posix emulation running on top of Windows
or vice-versa.
