Also, I appreciated this polemic:

One manifestation today (and then) is to take a likely good principle -- functional relationships can be powerful and understandable -- and then mess it up by wanting to use “old math function ideas” in the new domain in which both passage and time and the memory of state are both critical elements. This is the critical flaw in today’s “functional programming” enthusiasts.

(ง •̀_•́)ง them’s fightin’ words

I agree with with - FP is much of 'old math' - functions + data. Rather we want functions + identity + time - a new math for virtual objects in a computer. You can emulate identity with data in FP, but I think we want different models where identity is what we start with, then apply functions to produce new versions, while identity is preserved.

from orleans' docs

I think that the best way to model stateful systems using pure functional programming is still an open question. We need to continue developing new ideas and testing them. I think that Alan Kay is premature in declaring this to be dead end.

Yes, it just feels like it’s a caricature of functional programming. Pretty much every serious FP person I know understands this issue, and a lot of research is dedicated to overcoming it.

If you look at croquet or NAMOS or other things Kay has talked about, do you consider them FP? They do have a strong sense of 'consistent state' and 'functional transformation' from one state to the next.