I realized another possible spec and implementation issue with CSS transitions. I'll have to bring it up on www-style soon, although I'm not sure yet what I'm going to say there.
The question is what should happen (1) for CSS transitions on elements that have display:none at one end of the transition or the other, most importantly elements that are changing from display:none to some other value of display and (2) for CSS transitions on descendants of those elements, when the style changes are simultaneous with the display change. The easiest implementation definitely ignores (2), and probably (1) as well, although supporting (1) wouldn't be as hard as (2). In order to support (1) only, we would just need to initiate transitions on style changes when elements are coming out of the undisplayed map. To support (2), we'd need to keep style data for elements that are descendants of 'display:none' elements, which would require some substantive architectural changes, or at the very least, a significant expansion of the purpose of the undisplayed map (which currently exists only to contain elements on which style changes could cause the element to become displayed). However, if we want to skip transitions on elements that are coming from being 'display: none', also a real danger here that, if we want real interoperability here, we'll need to standardize the definition of what style changes are simultaneous. Skipping transitions on things that are display:none (rather than transitioning from it) also means computed style on subtrees inside display:none elements would be inconsistent with those outside; it seems like this could be a problem. It seems to me that the *right* thing to do is to make the architectural changes needed for transitions on elements that are or were inside display:none elements. However, that probably has significant performance costs: it would substantially reduce the performance saving of making elements display:none (including the cost of dynamic style changes that we currently skip processing for subtrees inside display:none elements). Then there's also the question of supporting transitions on elements that aren't in the document. I'm inclined not to care, although we do support computed style on elements not in the document (i.e., without a parentNode chain up to the document object). Thoughts? -David -- L. David Baron http://dbaron.org/ Mozilla Corporation http://www.mozilla.com/ _______________________________________________ dev-tech-layout mailing list [email protected] https://lists.mozilla.org/listinfo/dev-tech-layout
