I watched the John Resig presentation too and learned that CSS selectors always work from right to left. That would mean that doing this::
$('#foo p') Would extract all <p> tags and from that list subselect those who belong to #foo. Suppose you have 1000 <p> tags of them only 100 are inside #foo you'll have wasted 900 loops. Surely $('#foo') is the fastest lookup possible. Doing it this way will effectively limit the scope of the $('p') search and you will never be bothered about any <p> tags outside #foo. Or am I talking rubbish?