Re: Mouse Lock feedback

On Wed, Oct 26, 2011 at 11:40 AM, Dimitri Glazkov <dglazkov@chromium.org>
 wrote:

> Hi all!
>
> Vincent was kind enough to share his most excellent spec
> (http://dvcs.w3.org/hg/webevents/raw-file/default/mouse-lock.html). I
> have a few questions/suggestions:
>
> 1) The case of cross-origin frame isn't handled properly with
> document.mouseLocked. We probably shouldn't return an element from a
> cross-origin frame, but returning nothing seems bad, too. Perhaps we
> should separate the queries of "is there a mouse lock in effect?" and
> "which node has the lock"?.
>

Good point.

Pending additional feedback:
- I will modify the spec to have only a bool returned from the function to
answer "is there a mouse lock in effect?".
- I will not create a new method to answer "which node has the lock?"
because it was a convenience only. Pages needing to track that information
are not expected to be common and they will be able to do so by storing the
current element in a variable upon lock success callback.

2) "mouse" seems too narrow. There's already a perfectly nice
> "pointer" term, which is used in pointer-events, and it appears much
> more inclusive of the newfangled touch pads and touch screens.
>

Given point 3 below I see a case for this.  Previously we stayed with 'mouse
lock' because that was the common name for the technique.

>
> 3) Putting the API on the document is misleading. There's only one
> pointer on the device, and you aren't really locking things per
> document. Also, the document is already a huge pile of methods and
> properties. Perhaps instead we should use Navigator as the right
> place? Something like this:
>
> navigator.pointer.lock(element, failureCallback, successCallback);
> navigator.pointer.unlock();
> navigator.pointer.isLocked();
>
>
On document vs navigator I don't have a strong opinion. We can move to
navigator if it is more appropriate. Either way the spec would make clear
that the API would have impact operating system wide.

Nesting the api in a member such as .pointer. feels good in that it bundles
it, no objections there.

I'm curious if anyone else has any opinions here?