Michael Gratton wrote: > I tested it against Xerces-J 1.0.3 and 1.0.2 which both failed to output > the children and against 1.0.1 which seemed to work fine, but not > against the CVS version.
Okay, now I'm able to duplicate it. I *never* use NodeList to iterate over the children because it is inherently slow because the iterator must be "live", per the DOM Level 1 specification. Not too long we implemented a caching mechanism so that the performance of querying the children via the NodeList would be better. So I figured out that the cache was not being invalidated when the document was cloned. And since the main clone operation set the default values of the cache fields to zero, the NodeList thought that the length of the children was zero -- that there were NO children! (I hope you can follow this. If not, check the current code base.) Anyway, finding this problem has revealed that our cloneNode implementation across our DOM is pretty bad. Therefore, I'm working on going through it all and fixing all of the clone operations. I'll let you know when I'm finished. > ! "Scientific progress goes 'boink'?" Calvin and Hobbes RULES! :) -- Andy Clark * IBM, JTC - Silicon Valley * [EMAIL PROTECTED]
