safe vs. unsafe (Was: Haskell Platform proposal: Add the vector package)

[Simon Marlow]

On 20/07/2012 10:16, Wolfgang Jeltsch wrote:
> Am Freitag, den 13.07.2012, 10:49 +0100 schrieb Simon Marlow:
>> Here it is:
>>
>> http://community.haskell.org/~simonmar/papers/safe-haskell.pdf
>
> On page 5, the paper mentions that RULES pragmas in an imported module
> may change the meaning of the importing module. This is true. However,
> I see another problem: RULES pragmas in the imported module can make the
> code in the imported and the importing module non-deterministic, as it
> is not clear which rules are fired. Is there a reason for this problem
> not being mentioned in the paper? Or am I understanding something wrong?

RULEs are necessarily non-determinstic because they match on 
expressions, but that would only be observable if the RULE was not 
semantics-preserving.  So it is the fact that a RULE may be 
non-semantics-preserving that is the real problem, not the non-determinism.

Cheers,
	Simon