- From: Peter Moulder <peter.moulder@monash.edu>
- Date: Fri, 24 Sep 2010 22:56:39 +1000
- To: www-style@w3.org
My apologies, there was an error in the test case. The revised implementation results are that most browsers seem to ignore the 'direction' property of the element with 'clip' as far as how the 'clip' coordinates are interpreted. (I also tried setting direction:rtl on the root element in case any browsers were testing an ancestor; but I didn't check this variation for all browsers. I didn't check any other possible interpretations of "in right-to-left text".) Konqueror was the only one where I found rtl making a difference to 'clip' coordinates. It apparently tests the 'direction' property of the element that has 'clip', and (in the rtl case) positive offset means farther to the left, i.e. the behaviour that I considered more desirable rather than the behaviour that I considered the best reading of the text. Further evidence that this is the direction the spec author intended comes from the text's description of 'auto' for <left> in the right-to-left case. I suggest changing # specify offsets from the left border edge of the box in left-to-right text # and from the right border edge of the box in right-to-left text to | specify an offset from the left border edge of the box if that box's | 'direction' is 'ltr', | and a leftward offset from the right border edge of the box if that box's | 'direction' is 'rtl'. I also note that visuren.html#propdef-direction appears to be trying to give a complete list of the effects of the 'direction' property, but the list seems to omit 'clip', resolution of border conflicts with collapsed table border model, which margin is ignored (10.3), and probably other things (search for propdef-direction). I suggest either adding to its list of effects, or making it clearer that it's not a complete list. pjrm.
Received on Friday, 24 September 2010 12:57:10 UTC