On Sun, Jan 04, 2009 at 04:19:38PM +0800, Evan Laforge wrote:
If you don't have set-car!, then identity and equality are impossible to differentiate.
There's still eqv?. (I wish people wouldn't use eq? as an example of an identity-comparison operation. It's as underdefined as unsafePtrEq.) So although state implies identity, the converse is not true. You can also have immutable objects with distinct identities. Some dialects of Scheme have recently started leaning towards making pairs immutable, but whether they should also make them indistinguishable is a separate question: http://groups.google.com/group/comp.lang.scheme/browse_thread/thread/7eccba9... And having a limited form of observable identity has even been proposed for Haskell: http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.31.4053 Personally, I find the idea appealing (and have implemented the Refs in the paper with IORefs and unsafePerformIO), because really, even currently a programmer has to care about sharing since it can have radical implications for performance and memory usage. Making it observable in the program would just mean acknowledging the fact that in real-world programming, you can't _really_ replace a variable with its definition without changing its behavior in important ways. Lauri