It’s just data

IE8 Beta 7000 Bug

Start with this page.  View with IE8.0.7000.0 from the Windows7 beta, first without javascript.  Then apply Sjoerd Visscher’s tip:

// allow IE to recognize HTMl5 elements
if (!document.createElementNS) {
  document.createElement('article');
  document.createElement('aside');
  document.createElement('footer');
  document.createElement('header');
  document.createElement('nav');
  document.createElement('time');
}

Result.  Renders considerably better than any previous beta or released version of IE.  The nav bar and data dropdowns aren’t repositioned because IE8 doesn’t implement document.addEventListener("DOMContentLoaded", ...), which I can work-around if I chose to.  Scrolling further down the page reveals a different problem...

IE8 DOM

Looking at the DOM, you can see the issue: after the first two articles, subsequent articles are nested.  Note that the svg element appears to be processed into a single node without the need for Sjoerd’s workaround.  I’m presuming that that is a result of IE8’s so-called Improved Namespace Support.

But I said “appears” above.  If I expand the svg node, I can see the individual path elements.  But if I go back to my source and remove those path lines from the HTML, the enclosing article will close correctly; the error simply moves down to the next article.

I’m presuming that this is a bug.


Does the problem go away if you change <path /> to <path></path>? Assuming that IE8 is still parsing as plain old HTML, the /> shorthand has no meaning in HTML land, right?

That page is actually pretty broken for me in Opera, too. It looks like it too is interpreting <path /> as <path>, so everything after it in the document is a child of that element.

Posted by Martin Atkins at

It is true that Opera 9.63 does not implement the currently commented out HTML5 SVG proposal; but it is equally true that they aren’t advertising that they do.  If memory serves, Opera 10 will do better; WebKit already does today.

Microsoft, by contrast, has heralded what they call “Improved Namespace Support”.  At this point, I can’t tell whether it is not carefully thought out, or if the implementation has a few more bugs that need to be shaken out.

Expanding the SVG node in the example page indicates to me that the /> shorthand is being somewhat honored, otherwise I would have expected the path elements to be nested:

SVG expanded

But to answer your question, here is the result if there are separate open and close tags for each path element:

Paths explicitly closed

So, in both cases, the path elements are recognized as having no children, the svg node only contains the three paths, and the subsequent two p elements and one footer element are siblings of the svg node.  The one difference is that the </article> tag has no effect in the version where the path elements are using the self-closing syntax.

This I can’t explain.  Even weirder, this only occurs once on a page.  If you look at the DOM on the top of this post, all subsequent articles have the same nesting.  “fixing” one svg causes the next svg, but only that one, to cause its parent node to not properly close.

Posted by Sam Ruby at

Sam Ruby: IE8 Beta 7000 Bug

ふむー...

Excerpt from はてなブックマーク - Bookmayuk - お気に入り at

Here’s a simpler demonstration.

This xmlns handling is new in RC1 (i.e. post beta 2) releases. It seems to apply only to unrecognised elements (i.e. not the standard HTML ones), that have an ‘xmlns’ attribute with any non-zero-length value.

The issue you’re experiencing seems to be that self-closed elements inside the namespaced subtree result in an equal number of end tags after the subtree being ignored, which sounds like entirely unintentional behaviour and therefore a bug.

Posted by Philip Taylor at

I just got [link] to the point where it renders acceptably on IE8 beta 7000, Opera 9.63, Safari 3.2.1, Firefox 3.0.5, and Chrome 1.0.154.43.

WOOT!

Posted by Sam Ruby at

Redo eller ej - HTML5 är på tapeten just nu

Igårkväll satt jag och försökte ta reda på hur väl den senaste tidens ökande förekomst av fraser typ “du kan använda HTML5 redan nu” stämmer in på mina egna märkspråksbehov och önskemål.Ni som till äventyrs följer mig på Twitter......

Excerpt from En webbplats på svenska om webbstandard at

ie8 vs intertwingly

Dean Hachamovitch:  I always suspected this was the case, but I guess that confirms it:  I’m not a real person. Screenshot: Intertwingly on IE8. To be fair, that’s actually better than versions of IE before it.  It still... [more]

Trackback from Sam Ruby

at

Add your comment