Python has shared mutable state. Modifying the contents of one variable can cause another variable to change. For example,

>>> a=[1,2,3]
>>> b=a
>>> a[0]=17
>>> b
[17, 2, 3]

I think it is unnatural for the value of

b

to change when I modify

a

. I've seen lots of evidence on the internet that this is a source of confusion for novice programmers who are encountering Python for the first time.

This is indeed a gotcha, but for this conversation it's worth separating globals from aliasing. You'd have this problem even if Python had no globals or shared state. (Mutable state is a prerequisite, yes.)

I mean, Rust lets you share state as long as it's immutable - globals are just particularly aliasy by default, but local mutable state can be no less shared

In Curv, 1. Mutable variables have disjoint state. Assignment has copy semantics. Copy semantics are easier to understand, and they are one of the features I need to avoid having shared mutable state. And this avoids aliasing. 2. Arguments are passed by value, not by reference, and that's another feature that avoids aliasing. 3. When a lambda expression is evaluated, the values of non-local variables are captured in a closure, but they are captured by-value, not by-reference. This feature is closely related to by-value assignment and by-value parameter binding. You could describe this as another situation where aliasing is avoided, I suppose. These 3 features of Curv semantics account for the fact that there is no shared mutable state.

@Doug Moen what does "mutable variables have disjoint state" mean? Just "copying" instead of aliasing? (Quotes since copy semantics at its best copies less rather than more.) With 3, I suppose it's why Java (mind I haven't touched Java in years) required closed variables to be final. (Of course, since those variables easily store mutable objects... I never understood why the rigmarole though "rigmarole" seemed like Java's motto.)

Curv has no shared mutable state, so variables cannot share mutable state. Modify one mutable variable, and the contents of other variables are guaranteed not to change.