Josef Sveningsson wrote:

> The report doesn't even say that Haskell should be lazy, only that it's
> non-strict. Now, it happens that most (all?) implementations have very
> similar operational semantics, but I don't think that a Haskell library
> should assume anything more about the semantics other than that stated in
> the report.

The formulation in the report is deliberate.  It allows different implementation
techniques to be used.  There is (at least) one implementation, namely pH from
MIT, that uses parallel evaluation.  It does definitely not implement lazy evaluation,
but with a fair scheduler you can't tell the difference from just observing the
output.

--

        -- Lennart





Reply via email to