Lambda the Ultimate

Because internal calls to non-open methods are dispatched statically, developers can change the way these methods are called without affecting subclasses.

Not to infuriate the audience but to understand the point: how is C# (or even old plain C++) falling short of this?

And does not freeze in various mixin/open class calculi do exactly that - promising no changes in binding? Like this or this.

Or is the key phrase "subclasses cannot intercept internal calls and thus cannot become dependent on those implementation details"? Looks solvable by finalizing the method.

The last possibility is "By declaring a method 'open,' the author of a class is promising that any changes to the class will preserve the ways in which that method is called". The only problem with this (for me) is specifying these ways. As it seems that open methods are the only with dynamic dispatch, and the language (fix me?) does not enforce the author's promise, then I don't see why one cannot just say this whole idea is just a good practice to follow in any existing language with static dispatch being the default, instead of a new language feature.

Uh, sorry if spilling my coffee during the 7AM call made me a bit unreasonable. I will be glad to understand my mistake :-)

The way that Jonathan's technique differs from C#'s default or final methods is that it doesn't prohibit overriding in subclasses. You can -still- override a method in a subclass, and when an external client calls it, the method is dispatched dynamically in the usual fashion. However, within a class definition, any message sends to self are dispatched statically, unless the method is marked open (in which case it is always dispatched dynamically). This permits you to retain the flexibility of inheritance, but in a semantically sound way: that's why Jonathan was able to prove a parametricity result about his language. It's really simple and really clever.

I'm not familiar enough with mixin calculi to be able to compare, though.

Froelich notes that not all uses of inheritance can be easily expressed with delegation, but I'm not sure whether he means it's impossible or just hard, and I'm also unsure what he means when he says it's "just the dangerous" uses of inheritance that are hard to express. One thing's for certain, though -- I wouldn't even have considered making any serious use of inheritance before seeing Jonathan's idea.