393340 - createElementNS without namespace in HTML document creates a broken element

Status: RESOLVED WONTFIX

(Core :: DOM: Core & HTML, defect)

Description

[Jesse Ruderman]

Attachment: testcase

The following script in an HTML document:

  var newHR = document.createElementNS(hr.namespaceURI, hr.tagName);

creates an element that doesn't actually get displayed as an HR.  It looks the same in DOM Inspector (same tagName, same lack of namespaceURI) but behaves differently.

This breaks one method of serializing and reconstructing DOMs, where serializing means listing each element's namespaceURI and tagName and reconstructing means using document.createElementNS.  I have to do special things for elements in HTML documents that don't have a namespace, and even that doesn't quite behave consistently.

Comment 1

[Jesse Ruderman]

Note that I don't really care *what* namespace HTML elements have (when they were created by the text/html parser).  I just care about being able to use that namespace to construct new equivalent elements.

Comment 2

[Jonas Sicking (:sicking) No longer reading bugmail consistently]

Actually, per current specs, we're doing the right thing. .namespaceURI on HTML elements (not XHTML) is null. And calling createElementNS with a null namespace will create just a plain element, not a HTML one.

Comment 3

[Jesse Ruderman]

Where do you see that?  http://www.w3.org/TR/DOM-Level-2-Core/core.html#ID-2141741547 looks to me like it's saying that createElementNS with namespace null should be equivalent to createElement.

Comment 4

[Jonas Sicking (:sicking) No longer reading bugmail consistently]

Actually, according the the spec they are definitly different. createElement creates nodes with a null localName, createElementNS creates nodes with a non-null localName.

However we don't follow the spec on that and always create nodes with non-null localName.

A problem is that if we let htmlDoc.createElementNS('foo', null) creates a HTML element, there is no way to use a htmlDoc to create plain XML elements in the null namespace.

FWIW, the HTML5 spec says that HTML elements should have the xhtml namespace. But that is a bit weird too since then your call would create an XHTML element rather than a HTML one.

Comment 5

[Jesse Ruderman]

> But that is a bit weird too since then your call would create an XHTML element
> rather than a HTML one.

Will there be a difference?

Comment 6

[Jonas Sicking (:sicking) No longer reading bugmail consistently]

XHTML and HTML elements are different yes. The simplest difference is that one is case sensitive and one is not. But there are also a bunch of web-compat-quirks that we apply to HTML elements that we don't on XHTML elements.

Comment 7

[Jesse Ruderman]

Ugh.  Is there some way for me to tell whether an element is an "HTML element" rather than an element that happens to have a null namespace but is not an HTML element?

Comment 8

[Alex Vincent [:WeirdAl]]

foo instanceof Components.interfaces.nsIDOMHTMLElement?

Comment 9

[Jonas Sicking (:sicking) No longer reading bugmail consistently]

myElement instanceof HTMLElement

Comment 10

[Henri Sivonen (:hsivonen)]

Using createElementNS with the XHTML ns URI works for me even when the DOM is in the HTML mode.

Comment 11

[Jesse Ruderman]

Attachment: testcase using localName instead of tagName

This testcase works now, thanks to the fix for bug 468708.  The original one doesn't because tagName is uppercase.

Comment 12

[Jesse Ruderman]

That makes me happy.