Most significant paragraph from the whole piece:<p><i>Though YUI performs ably, it’s my opinion that the numbers seen in TaskSpeed should be taken with a hefty grain of salt. The tests are designed to exercise library abstraction logic against DOM-intensive operations. The issue here is twofold:<p>1. Not all libraries (YUI included) have abstraction logic for all of the specific tasks, which breaks the apples-to-apples comparison.<p>2. And in order to get meaningful numbers, the test operations are iterated up to 500 times or performed against excessive numbers of nodes. In real-world cases, these conditions are not the norm, meaning the differences are exaggerated, perhaps even grossly.</i>