On Mon, 13 Sep 2010 14:54:14 +0200, Henri Sivonen <hsivonen@iki.fi> wrote: > Then there's the issue of whitespace-only text nodes uselessly > reconstructing the list of active formatting elements. I think changing > this would be potentially more disruptive in terms of the code and test > cases involved. Also, making the most common tree builder mode ("in > body") sensitive to space characters vs. non-space characters would be a > performance problem if implemented naively not optimized properly (to go > on the fast track once one non-space character has been seen until a > non-character token is seen). Still, having seen a page like > http://saintjohnchurchmiddletown.com/default.aspx (warning! may slow > down your browsing considerably while loading) in the wild, I have to > wonder if we still should change the spec not to reconstruct the list of > active formatting elements on space characters. I think this would be a good change. Old Gecko didn't reformat for line breaks. When I suggested changing the spec before to not reformat for whitespace, Hixie said the spec is as it is for compat, but didn't have a list of URLs that would break. It seems the above URL works fine in Opera but is considerably slower in Firefox and Chrome with the HTML5 parser. -- Simon Pieters Opera SoftwareReceived on Monday, 13 September 2010 16:18:48 UTC
This archive was generated by hypermail 2.3.1 : Thursday, 29 October 2015 10:16:04 UTC