Re: [geometry] DOMQuad is weird

On Wed, Oct 19, 2016 at 8:39 PM, Simon Fraser <smfr@me.com> wrote:

> > On Oct 19, 2016, at 3:02 pm, Tab Atkins Jr. <jackalmage@gmail.com>
> wrote:
> >
> > On Wed, Oct 19, 2016 at 12:00 PM, Simon Fraser <smfr@me.com> wrote:
> >> As a naive JavaScript author, I would consider a point to be a primitive
> >> type that's copied by value.
> >>
> >> Dean points out that the primitive types are actually DOMPoint.x and
> >> DOMPoint.y, so what I wrote above is "expected" JS behavior.
> >
> > Right. Given the attribute-based API, it's generally expected JS
> > behavior that you can detach an object, alter it, and the parent
> > object (which is still referencing it) will see that change.  It's the
> > same behavior you'd get from the original object you passed in to the
> > constructor.
> >
> >> Maybe DOMQuad should be entirely read-only, or should return copies
> >> of points (getP1(), getP2() etc).
> >
> > That would indeed communicate that the points are likely to be fresh
> > objects, but it's a rather non-idiomatic thing for such simple
> > objects.  Readonly is possible, but that precludes some use-cases.
>
> Why not follow DOMPoint and DOMRect and have read-only and mutable
> variants?
>

Doesn't that seem a bit like overkill?
Are there places in the platform where an assignment of an attribute
creates a copy? I know that there's a desire to not have live objects that
point to something on the native side but that doesn't apply here.

Also, Firefox is already shipping DOMQuad with this behavior.