Basic: I use PHP5.4.9 from Ubuntu 13.04 repository. The aim of my code is to iterate HTML source (as a DomDocument) recursively and cleanup everything that is not valid to be used inside Epub files (i.e. and attribute align is not valid for paragraphs in Epubs).

FIRST: Today I tried removing attributes from a DOMElement using this simple code:
<?php
for ( $k=0; $k < $element->attributes->length; $k++) {
    if( /* some rule */ ){
        var_dump( $element->attributes->item($k)->nodeName);
        $element->removeAttributeNode( $element->attributes->item($k));
    }
}
?>
Unfortunately all attributes still existed when this loop was finished, even if these var_dumps told me that deleting them was tried.

I solved this problem iterating the attributes list backward:
<?php
for ( $k = $element->attributes->length - 1; $k >= 0; --$k) {
    if( /* same rule */ ){
        var_dump( $element->attributes->item($k)->nodeName);
        $element->removeAttributeNode( $element->attributes->item($k));
    }
}
?>
SECOND: DOMElement::removeAttributeNode does NOT return a bool but a DOMAttr object.