On 8/13/07, Simetrical Simetrical+wikilist@gmail.com wrote: [snip]
If there is in fact a performance decrease in more extreme but not ridiculous cases, then of course it should be fixed even if the result is a bit ugly.
Not doing anything on this front doesn't bother me.
I've been able to create a noticeable performance reduction of the anchor scanning approach but only in a contrived test case. Then again I am testing on fairly fast client hardware.
I will, however, be calling on your defense should someone reject some nifty tool in the basis that scanning anchor tags might theoretically be slow. ;)
(Iteration through DOM objects matching classes was cited as a performance concern for the DIV order randomization hack I made for boardvote on meta.)