On Mon, Feb 23, 2015 at 6:31 PM, Keean Schupke <[email protected]> wrote:
> On 23 February 2015 at 23:05, Matt Oliveri <[email protected]> wrote:
>> To me, you've stretched the usual idea of subtyping because you seem
>> to effectively require specialization and/or inlining to avoid the
>> coercions that come with naive subsumption. This is what we were
>> consistently miscommunicating about before though, so maybe you don't
>> agree.
>
> Subtyping is a type-system only relation, and it has nothing to do with
> whatever process the compiler is actually doing. All we require of the type
> relation is that it only allows types that the compiler will produce correct
> results for (soundness). That is all, it doesn't matter if the compiler is
> specialising, generalising, or performing some other strange operation.

OK, fine. So you are stretching the usual set of implementation
techniques for subtyping.
_______________________________________________
bitc-dev mailing list
[email protected]
http://www.coyotos.org/mailman/listinfo/bitc-dev

Reply via email to