[Grace-core] Minutes of Teleconference 2011.08.17

[James Noble]

>> * don't need comprises on class declarations - just use explicit unions?
> 
> I'm missing all of the context, of course, so feel free to ignore this.  But I don't get this.
>  I thought that we had been using comprises on TYPE declarations to say that here we want 

short answer: yes. 

longer answer: we'd be talking about a comprises annotation on (abstract super) classes or types. 
we now think union type declarations are probably enough, so we don't need both. 

> one of the named classes rather than any structurally-matching class.

even longer answer: distinguishing structurally equivalent classes really turns out to need BRANDS.

I was telling myself we're almost done with pattern matching now  -
 or at least I've mostly convinced myself (again) that I can see through to the end of it

but we do need also to fill in that gap, have some notion of brands and how they work.

Or else people will start writing methods like "i_am_a_FOO" to match structurally..

J