No one said that .children is being "de-promoted". All I said is that there is no child node array any more, so retrieving a child is potentially O(N). Surely for small N, you won't see a difference. If you want to contribute some real insight here, please try with 600 or 6000 children (or post the test for others to try), doing timing on 6 children proves nothing. If for 6000 children .children[1] is still as fast as getting the next sibling, then you won indeed. While you're there, compare getting the 1st and the 6000th child. Maybe retrieval is really smart and the linked list of children is only walked to the number you require.
Bug 1535725 Comment 49 Edit History
Note: The actual edited comment in the bug view page will always show the original commenter’s name and original timestamp.
No one said that .children is being "de-promoted". All I said is that there is no child node array any more, so retrieving a child is potentially O(N). Surely for small N, you won't see a difference. If you want to contribute some real insight here, please try with 600 or 6000 children (or post the test for others to try), doing timing on 6 children proves nothing. If for 6000 children .children[1] is still as fast as getting the next sibling, then you won indeed. While you're there, compare getting the 1st and the 6000th child, and also the 5950th, maybe there is optimisation to get the last one. Maybe retrieval is really smart and the linked list of children is only walked to the number you require.