- From: Johnny Stenback <jst@netscape.com>
- Date: Tue, 22 Jul 2003 17:07:22 -0700
- To: Christian Parpart <cparpart@surakware.net>
- CC: www-dom@w3.org
Christian Parpart wrote: > -----BEGIN PGP SIGNED MESSAGE----- > Hash: SHA1 > > hi all, > ... > 2. interface CustomEvent > > I do not believe that CustomEvent is needed since I even just implemented it > after have the main stuff working from the events stuff. > AFAIK CustomEvent gives just access to some members for at least the > EventTarget's dispatching methods. While I do not know whether Java does > support friend classes to access private members I still propose to remove > this interface since it is very implementation specific and there is really > no need for the application writer to have access to these private members of > the Event interface - or give me some reasonable examples :o) One scenario where this is needed is in mixed language implementations like Mozilla (just to name one). There script writer can create custom events in JavaScript (or whatever supported scripting language, at least in theory) and dispatch them, in this case the C++ implementation that does the actual dispatch needs to be able to set the target, and so on. In this case it gets really hard if not impossible to do this w/o a well-defined interface on custom events that have setters for those readonly attributes in question. > 4. EventFOO.initFOO[NS](...) > > Letting all these methods returning the Event that they are (this) would be > very helpful for the application programmer since it sometimes makes sense to > combine several functionis together. > > There's my C++ use case: > > dispatchEvent(document->createEvent("Foo")->initFoo("bar")); > > Although while C++ does not have any GC I assume that the > EventTarget.dispatchEvent() method may take ownerchip of its > passed event object. does this violate the spec to be conform? This is more a coding style issue than anything else. As for the ownership issue, the spec doesn't attempt to define that since there's no common rules that apply in every case. IOW, it's up to the implementation to define the ownership of DOM objects. Making the DOM implementation delete events that are dispatched is IMO not a good idea since it limits what type of event objects you can pass in. For instance, you can't create a custom event on the stack (i.e. w/o allocating it on the heap) if your dispatch code deletes the events after they're dispatched, and you can't create an array of events and pass in an event in that array (since the array items can't be deleted, all arrays might live in one contigous block of memory). And so on. ... > > That's it so far, > Christian. > > - -- > 15:59:54 up 62 days, 7:08, 0 users, load average: 0.00, 0.00, 0.00 > -----BEGIN PGP SIGNATURE----- > Version: GnuPG v1.2.2 (GNU/Linux) > > iD8DBQE/G/bbPpa2GmDVhK0RAuztAJ9BuC2cCW/P7l0P+E/CKHzgtgvHqwCfWt3h > vBe8ugKZRVUrtwzSFuPanw0= > =G0e5 > -----END PGP SIGNATURE----- > -- jst
Received on Tuesday, 22 July 2003 20:08:20 UTC