Echo Web Framework

I just installed IE8RC1 and tried it out against the Echo framework. The first test of course was to point it at our main demo application (http://demo.nextapp.com). It downloaded the application's scripts and rendered the initial screen reasonably quickly. But then I tried dragging the "Welcome" window of the demo app, and it seemed a bit slow. Opening the task manager, I also noticed that *any* mouse movement would peg the CPU at 100%, even over elements that had no mouse listeners registered.

Then it got worse. I clicked the "Next" button at the bottom of the screen to show the next feature demo, a tabbed pane with a few windows on top. It was not a particularly complex document model, but a bit more so than the previously loaded "Welcome" screen. The browser took several seconds to load it, and I additionally noticed that even mouse rollover effects became sluggish.

Internet Explorer 8 consuming 100% CPU while trying to update the screen after rendering 21 nested DIVs. Note the artifacts from dragging the task manager window over the IE8 window.

Then it became simply unusable. I clicked the "Next" button again. The next test is the "SplitPane" demo, which divides the screen up into eight resizable panes. It took an entire minute before it displayed. When it did display, the screen slowly rendered from top-to-bottom like it was an image being downloaded on a 2400 baud modem. I rolled the mouse over the resize handle of one of the rendered SplitPane components. IE8 RC1 took ten seconds before the mouse cursor changed from a pointer to a resize arrow, and another ten before the resize handle's rollover effect image displayed.

After a few attempts to replicate the problem separately from Echo3, I eventually thought to create a static HTML page containing a dozen or two nested and absolutely positioned DIV elements. It turned out that nesting 17 absolutely positioned DIVs inside one another caused the IE8 browser to enter a state of relative panic. You could watch the outer DIVs render rapidly, and then see IE gradually slow down as it rendered each successive inward DIV element. The performance slowdown appears to be exponentially proportionate to the number of nested DIVs. 25 nested DIVs took minutes to render.

My static HTML test case can be found here: http://echo.nextapp.com/content/test/ie8/

I tried demos from other AJAX frameworks as well. Qooxdoo's "Demo Browser"--which can render some moderately complex UIs--seemed to have great difficultly with IE8. Other AJAX toolkits whose demos tended to only show off one widget at a time (resulting in a less complex/nested DOM) had better luck.

"Normal" web sites work reasonably well, as long as they don't employ many nested absolutely positioned DIVs. But absolute positioning is just about the most fundamental aspect of rendering an application-style user interface within a web browser. Most any complex AJAX application UI simply will not work in this version of Microsoft's next-generation browser.

I can't fathom how this shipped as a "release candidate". My concern of course is whether this problem will ship in IE8 final. I've reported a bug, and Microsoft has already responded. Unfortunately the response so far has been "can-not-duplicate":

http://www.microsoft.com/communities/newsgroups/list/en-us/default.aspx?...

Microsoft stated the following on its IEBlog Release Candidate Announcement:

I hope (and believe it likely that) this will be treated as a critical issue.