There's a second commit that simplifies the specification of advance(I, S) which became too complicated post-Kona. My Assignable changes make the requirements worse, or at least no better, this change is crystal clear.

I'm not sure how I feel about this. Think of output iterators and proxy references. In what sense is some value assignable to *i when i is an ostream iterator? I feel like Assignable isn't the right place to put these semantics.

We don't use Assignable in Writable and MoveWritable for exactly that reason; the semantics are different.

EDIT: I've effectively co-opted Assignable from being only syntax to participating in the Object concept family.

EDIT 2: I've also used "equals" instead of == here as a quasi escape clause. A value is assignable to *i for an ostream iterator when the conceptual object it denotes in the output sequence is "equal" to the value assigned. E.g., *i = 42 should insert "42" into the output sequence, not "37". When T is StreamInsertable, which it must be for us to assign it through an ostream_iterator, we can construct a common type for T and string that validates that relationship.

Integrating these semantics for Assignable consistently with iterators would be in scope for https://github.com/CaseyCarter/stl2/issues/17, and definitely something I'll be thinking about at length.

There's a fresh pull with Assignable requiring Common. I've yanked the second commit with the advance changes from the branch since they are a separate issue, and I'm working on a more comprehensive cleanup of Sentinel / SizedIteratorRange / advance.

I'm not seeing the Common requirement of Assignable. Did you forget to add it?

I'm not seeing the Common requirement of Assignable. Did you forget to add it?

I reset the branch, and ran git commit --amend without adding the changes to the index. Yay me.