#rust-lang

FYI: wrote up thoughts about specialization, coherence, and the path to stabilization -- http://aturon.github.io/blog/2017/02/06/special...

cc woboats

aturon: I think you missed a #[fundamental] on ParentTrait in the `The fundamental attribute` sectin?

*section

sfackler: no, the use here is about fundamental applied to types

(like Box)

oh got it

so iiuc, we're stuck with #[fundamental] forever? and with that settled, basic specialization is "done" design-wise?

scott: also I was talking to wycats about the notation and IT hink we should schedule a kind of "focused call" to discuss it

sure

Ixrec: that’s accurate

well there’s some lifetime stuff with specialization that i don’t think is resolved, of course

still sounds like progress at least

I would like to see if we could get it into FCP in the 1.18 or 1.19 cycle but I may be optimistic

yeeeeees

so what's the unresolved "lifetime stuff"?

Ixrec: specializing on lifetimes is unsound

yeah, I thought it was as simple as "we won't let you do that"

I think that’s where we’re at for now at least

but it has surprising outcomes actually

for example, `impl<T> Foo for T` and `impl<‘a> Foo for (&’a str, &’a str)`

that’s specializing on lifetimes, because `(&’a str, &’a str)` is more specialized than `(&’a str, &’b str)`

however, `impl Foo &’a str` is not specializing on lifetimes; its specifically the fact that you’ve said two lifetimes are equivalent

which of course usually doesn’t matter