[whatwg] Proposal for public data in drag events

[Trevor Burnham]

On Aug 30, 2012, at 1:27 PM, Ian Hickson <ian at hixie.ch> wrote:

> On Thu, 21 Jun 2012, Trevor Burnham wrote:
>> 
>> I've been using HTML drag-and-drop 
>> (http://www.whatwg.org/specs/web-apps/current-work/multipage/dnd.html) 
>> in a project, but I've run into one limitation that seems severe to me: 
>> There is no direct way to determine what the source node is from a 
>> dragenter, dragover, or dragleave event. This makes it difficult to 
>> support use cases where elements react to those events differently 
>> depending on what is being dragged over them.
> 
> This is intentional, because that source node could be from a Web page in 
> another origin, another browser, or indeed, an app that isn't even a 
> browser. So there's no real sane way to do it.
> 
> 
>> I understand that the reason for this is cross-document drags: In 
>> addition to security implications, obtaining a reference to a DOM node 
>> in another document simply wouldn't make sense. Therefore, the 
>> dataTransfer object only allows serialized data. Unfortunately, 
>> dataTransfer is only appropriate for carrying data to the drop target. 
>> There is no mechanism for providing data to intermediate drag event 
>> receivers, except for the "types" attributes on the dataTransfer object. 
>> "types" can be used to carry data that you want to make public 
>> (http://stackoverflow.com/a/11089592/66226), but this is clearly a hack 
>> and it carries some limitations. Most notably, the spec requires that 
>> data type strings be converted to ASCII lowercase.
>> 
>> Therefore, I'd like to propose the addition of a "publicData" object on 
>> all drag events. It would have the same interface and behavior as the 
>> dataTransfer object, with the sole exception that it would be read-only 
>> in all events where dataTransfer is protected. That is, publicData would 
>> be read/write in dragStart, and read-only in all other drag-and-drop 
>> events.
> 
> That's an interesting idea. I suppose we could expose it using a custom 
> type in cross-app OS dnd situations, too.
> 
> Could you elaborate on your use case? Are there cross-window use cases for 
> this? (For in-window cases, you could instead just use a global.)
> 
> -- 
> Ian Hickson               U+1047E                )\._.,--....,'``.    fL
> http://ln.hixie.ch/       U+263A                /,   _.. \   _\  ;`._ ,.
> Things that are impossible just take longer.   `._.-(,_..'--(,_..'`-.;.'

The main use case I have in mind is an interface where elements respond to the object being dragged. A common case is showing visual feedback depending on whether the element emitting a dragenter/dragover is a valid drop target for the object being dragged. Currently, this can only be done in response to the types attribute.

Using global state in tandem with the types attribute is a viable workaround for drags within a document, but an inelegant one. It's also incompatible with multi-touch. Although no multi-touch implementation of the dnd spec currently exists (to my knowledge), the spec does not preclude multiple simultaneous drags, so this may be a concern in the future.

Trevor Burnham
@trevorburnham