W3C home > Mailing lists > Public > www-dom@w3.org > April to June 1998

RE: NodeIterators & Java implementation

From: Alfie Kirkpatrick <akirkpatrick@ims-global.com>
Date: Mon, 11 May 1998 10:15:44 +0000
Message-Id: <TFSIFIUB@ims-global.com>>
To: www-dom@w3.org, donpark@quake.net
Sorry to jump in on this one. The case for COM is actually very
simple. Normally, when returning an object, the reference count
is increased and it is up to the calling function to decrement it
when its finished with the object. In the case of Node.removeChild,
the reference count would just be left alone, so that when the
calling function was finished with it, it would decrement the reference
count to zero and the Node would be deleted. I found memory
management to be one of the simpler things when implementing
the previous DOM draft in COM, provided you are careful and
follow the rules with AddRef and Release...

Would it not be possible for a Java implementation to mimic this
COM behaviour to get around the "GC problem". I don't understand
the Java case too well, so maybe this isn't an option. I suppose
what you (Don) are asking for is exactly this, written into the spec.
However, it might actually cause confusion in a COM implementation
(DOM release on top of the COM release!)


From:  donpark@quake.net
Sent:  08 May 1998 11:41
To:  www-dom@w3.org
Subject:  Re: NodeIterators & Java implementation

One topic that is missing from the spec is how to free/delete/release
objects obtained through DOM.  For example, what should one do with Node
returned by Node.removeChild?  Document interface has methods for   
objects but nothing to delete them.  Are we do assume that language   
facilities are to be used?  Do we use 'delete' operator or 'free()' under
C++?  What about CORBA and COM-based implementations?  One does not just
'delete' under COM.

Also how do we know which objects should be deleted and which one should   
be?  What is wrong is designating 'release' on Node as well as iterators   
Java specific delete facility?  For that matter, why not move it all the   
up so everyone can standardize on the delete facility?  It would certain
remove a lot of headaches.  If one is smart enough to invoke   
why is requiring invokation of release() too much?

Don Park
Received on Monday, 11 May 1998 05:29:36 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 20 October 2015 10:46:04 UTC