[Cuis-dev] Fake polymorphism (was: [DEFECT] #copyFrom:count: for OrderedCollections)

[Boris Shingarov]

On 2/6/24 17:24, Hernán Wilkinson via Cuis-dev wrote:

> ok. Which definition do you use?
> Could you also elaborate why you see it that restrictive?

Reminds me of our conversation in Tigre in 2017, where Theo called it 
"Exploratory polymorphism", which I see as opposite to what Andrés seems 
to have in mind when saying

> the message itself is wrong if it 

If I understood correctly what Theo was saying, in a "pure exploratory" 
universe such as Smalltalk, the word "wrong" just isn't part of the 
vocabulary.  You throw a message at an object and you observe what 
happens — your are not entitled to demand the object behaves in a 
particular way.  A toddler explores the world by interacting with 
objects.  She shakes a beanbag toy, and the toy makes a rattling sound. 
She then shakes a tennis ball, and the ball makes no sound.  She shakes 
her cat, and the cat scratches her.  *This* is exploratory polymorphism.

To use words such as "correct" or "wrong" meaningfully, we would need 
something else, something like ML's "modularity" or Haskell's 
"typeclasses", where "signatures" are constraints expressing the concept 
of "correct", see
https://people.mpi-sws.org/~dreyer/papers/mtc/main-long.pdf

In a live-model-of-universe, where objects are independent 
"computational actors", what people have been trying to model "correct 
behavior of the actor" is game-theoretic contracts, where you can 
construct proofs-of-correctness as in, "I can prove that if the objects 
I send messages from inside my method, satisfy the contract, then my 
method will satisfy the contract".