On 8/30/2011 8:31 AM, bearophile wrote: > Time ago I have suggested a "trusted purity" to allow the implementation of > pure function with memoization, but this observational purity seems better. "observational purity" seems like another word for "logical const". This has been debated here many times.