On Tue, Jun 16, 2015 at 1:06 PM, Matt Oliveri <[email protected]> wrote: > On Tue, Jun 16, 2015 at 7:05 AM, Matt Rice <[email protected]> wrote: >> On Tue, Jun 16, 2015 at 1:04 AM, Matt Oliveri <[email protected]> wrote: >>> On Tue, Jun 16, 2015 at 2:52 AM, Matt Rice <[email protected]> wrote:
(reordering a bit to reply to both these things below) >> and frustratingly set >> up so that you cannot speak about type's return value, or specify some >> return values future type existentially, > > I don't understand. A type doesn't have a return value. And what is a > "future type"? You're not talking about futures, are you? In a sense yes, parsers and languages are *full* of stuff which will only be representable in the future. value:type is one of these things, so we collect all the arguments *before* we can solve whatever it is we are trying prove, it is useful if everything is consistent across phases. >> its easy to fall under the spell of omniscient type values, rather >> than isolating phases from one another, and reusing the language. > > I don't know what you mean by "omniscient". But it kind of sounds like > you're against having code be reusable across phases. That is my main problem with type, is exactly that it doesn't have a return value, the low-level sequence is value:type = sequence_of_one_or_more_things, so there is a 1 < - n, (one type for every typed sequence), further there is a 1 -> n fields to a type (or we can go from 1 type, to n typed values), this expansion and contraction, is rather fundamental, and so saying that type doesn't have a return value is implying type is global and ambient, because it needs to be globally consistent, I believe there are other ways of proving global consistentcy, through isolation and the interpretation of the types value... this *allows* types to be reusable across phases, if we take away the "return value" of 'type', when we absolutely *remove* the ability to pass them as arguments and use parameter passing as a mechanism for giving or denying authority to types, so the easiest analogy I can come up with is that an ambient/global idea of type, essentially builds a fence, where it seems reasonable that a wall would suffice, you'll most certainly have to drop subtyping though, or at move to its equivalent in the space where types are values... will try and get to the rest, but in the middle of painting the house, hopefully this covers the main point i am trying to make. _______________________________________________ bitc-dev mailing list [email protected] http://www.coyotos.org/mailman/listinfo/bitc-dev
