W3C home > Mailing lists > Public > www-dom@w3.org > July to September 2002

Mixed level 1 and level 2 nodes woes

From: Sander Bos <sander@x-hive.com>
Date: Wed, 17 Jul 2002 16:31:00 +0200
Message-ID: <41D11F414A26E942912B7E7696DC8E220615ED@JAKARTA.xhive.archipel>
To: "WWW DOM" <www-dom@w3.org>

Hi there,

I have some questions about documents with mixed level 1 and level 2

My original question was:
  You have NamedNodeMap::setNamedItemNS(Node) and
Element::setAttributeNodeNS(Attr), which both say that when called it
should replace any node with the same local-name/ namespace URI pair. My
question is, what should happen when this method is called with a DOM
Level 1 node as argument, where both the local name and namespace are
null? I do not see any behavior defined for this, I see two options:
 - If namespace and local name are null on the entered node, call
setNamedItem/ setAttributeNode instead.
 - Throw an exception.

Then I thought about it a little more, what should happen if a user does
  Element myElement = document.createElementNS(null, "someName");
  myElement.setAttribute("sameName", "value1");
  myElement.setAttributeNS(null, "sameName", "value2");
  Element myElement = document.createElementNS(null, "someName");
  NamedNodeMap attributes = myElement.getAttributes();
In both cases, all statements individually are correct, and in both
cases the set...NS statement per specification only looks whether there
is already an attribute with the same localname. So with our DOM
implementation I get two attributes with the same name on the element (I
also tried it with Xerces, and there the attribute gets replaced with
the second one in both cases, so maybe I read the specification

I already know the standard answer: "Don't mix the two kinds of nodes
That is very nice, but users are going to do that if you give them a
chance, and there are at least two use cases where I cannot blame them:
 - If you have one level 1 document and one level 2 document, and copy
some data from one to the other using importNode. The specification says
you have to make copies of the namespace attributes (attributes of the
Node here), so if you import a DOM level 1 node, it will remain a DOM
level 1 node when copied (Our customer who ran into this was
constructing a document from several others).
 - If the user has first parsed a document (which will usually create
DOM level 2 documents by default in most implementations), and then
start to alter that document using DOM calls, but without an intimate
knowledge of the DOM specification. He/ she will most likely go for the
createElement, createAttribute, setAttribute calls instead of
createElementNS(null, ...) etc. if he/ she uses no namespaces, leading
to mixed-nodes documents. All the non-NS methods have the warning
	"To create an element with a qualified name 
	and namespace URI, use the createElementNS method."
which I do not think is going to stop most people, who will think 'I am
not using namespaces, I do not even know what the heck a qualified name
is, so createElement is the call for me'.

I also have a proposed solution: I feel that
 - createElement, createAttribute and setAttribute should create DOM
level 2 nodes when the document element of the document is a level 2
node (so localName should be set in that case).
 - importNode should also set the localName if a DOM level 1 node is
imported into a DOM level 2 node (do not really know for the other
direction, simply clearing the info is probably not possible).
 - setNamedItemNS and setAttributeNodeNS should not only check for
localName and namespace URI but also for qualified name, like Xerces
apparently already does.

Sure, it breaks backwards compatibility, but if most people don't care
about mixed node problems, they probably won't care about backwards
mismatches in those things.

Kind regards,


P.S.: Recipe to get two attributes with the same name on one element in
Xerces 2.0.1:
  Element myElement = document.createElementNS(null, "someName");
  org.w3c.dom.NamedNodeMap attributes = myElement.getAttributes();
will serialize to:
	<someName otherName="" otherName=""/>

X-Hive Corporation (www.x-hive.com)
email: sander@x-hive.com
phone: +31 10 7108625
Received on Wednesday, 17 July 2002 10:31:37 UTC

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