Re: [gtk-list] Re: drag and drop protocol: compatible with gtk?

[Elliot Lee <sopwith cuc edu>]

On 4 Jan 1998, Owen Taylor wrote:

> Great! Just what we need, another incompatible DND protocol.  While I
> don't think JX compatibility is very important, compatability with Qt's
> solution is crucial. (From Arnt Gulbrandsen's response, it sounds a bit
> like they might by going with the JX solution) 

The JX post notes that it is implemented in both JX and Qt, although
probably that Qt version is not yet released.

> I'll prepare a detailed critique of both protocols. Though it would
> be a lot easier if there was an actual description of the GTK
> protocol. The code leaves a number of unanswered questions, e.g:
> 
>  When are XdeEnter events sent to the target app? (The code sends
>  them on every motion event, that doesn't leave the drop zone, but I suspect
>  that is a bug)

Well, it used to send them on every motion event, but that was a bug(tm)
and it should work fine in the latest gtk.

The XdeEnter event is sent either when the mouse pointer enters a new
window, or when it leaves the drop zone in the present window.

>  What is the function of the 'sendreply' flag for XdeEnter and XdeRequest?

To theoretically allow use of the protocol for a clipboard protocol... 
I.e. in a clipboard protocol you'd send an XdeRequest with "sendreply" set
to 1, whereas in the D&D protocol you send XdeEnter with "sendreply" set
to 1. It's not used in the current protocol version, though. 

> > The Gtk protocol does things the way I think they should be done. The JX
> > protocol has some nice things, but it also has some real shortcomings as I
> > see it, and I don't think it is wise to support it.
>  
> Perhaps not. But I can't see the point in using a GTK-specific protocol.

Yes, compatability is important, but we either use a protocol that fits
our needs and does it "right", or we go chasing after the mythical god of
compatibility. People have already tried that strategy by emulating
Microsoft instead of creating the best possible solution. They've failed. 

> A subset of Motif DND could possibly be shoe-horned into the current
> API. (The fact that Motif uses selections presents a problem problems
> both for data transfers during the drag and after the drop.)
> But, I think things could be improved by presenting a more generic
> interface to the program.

Sounds cool, looking forward to it,
-- Elliot					http://www.redhat.com/
"They don't let my code go into shipping products," Gates said. "They
 haven't done that for eight years." (at the 1997 PDC)