François Pottier

> * The notion of immutable arrays is useful to enable compiler optimizations Indeed, and in particular, CSE: if the array `a` is immutable, then multiple occurrences of `a.(i)` can...

> * The implementation choice for immutable arrays is critical. Of course, > you do need as core building block an implementation of an array as a > single block...

Regarding the distinction between "truly immutable arrays" (immutable for everyone) and "temporarily immutable arrays" (immutable for me, but possibly mutable for someone else; or immutable now, but possibly mutable later),...

> This reminds me of the joke about searching for keys under the street light. Point taken, but there is something to be said in favor of searching under the...

Hi! I have read the discussion and will now post a few responses (in separate messages, for clarity).

@avsm writes: *it is occasionally more ergonomic to use the low level syntax when the existential type variables introduced by the pattern matching need to be accessed*. I believe he...

@talex5 makes the point that one must scan an entire `match` construct to check whether there is an `effect` clause, and that the presence of such a clause changes the...

@lpw25 writes: *this makes the default form of handlers deep handlers.* This may be true in the current proposal, but I do not see why it has to be so....

Regarding the separation of "operations" and "effects", @lpw25 writes: *This division gives a lot of additional flexibility*. While I am not opposed to this proposal, I note that a similar...

I think I am generally in favor of the distinction between names and operations; it sounds reasonable and attractive indeed.