MarkNicholls

In a sense you can already implement intersection types by implementing a type with both the constituent parts. You can also specifiy that a type is a member of an...

I've returned to this (because irritatingly I wanted to do the same thing again and couldn't). My only issue with this proposal is its name, we don't actually need intersection...

'impure' is quite common for this sort of thing (assuming I've understood) in contract proving frameworks.

I don't completely understand the purpose of this conversation, and what transient's purpose is (I assume as a marker to the runtime/compiler to aid optimisations?), but the issue of pure...

I don't think I'm in a position to give concrete suggestions, I am in a position to (naively) present how this is handled in specific contexts where it has been...

If you control the IO API (or you can specify how it should be 'declared' by implementors) then really all you need to do is to have a mechanism to...

I THINK its the same notion, I **think** you're grappling with what in my programming circles is called referential transparency/purity. I'll try to see if I understand the comments tomorrow.

@ChristianGruen I'm not sure i understand the nuances of the discussion, the principle in FP is all IO is impure, the issue really is you have to err on the...

I wont get bogged down in words, I'll use 'deterministic' for the moment (for me purity and referential transparency is synonymous, and determinism is slightly weaker). Asynchronous in my head...

> True. As basically all functions of the current XQFO standard are deterministic anyway, the challenge is how to make them nondeterministic. in terms of a specification, naively you just...