- From: David Hyatt <hyatt@apple.com>
- Date: Sat, 26 May 2007 21:08:30 -0700
- To: Preston L. Bannister <preston@bannister.us>
- Cc: Daniel Glazman <daniel.glazman@disruptive-innovations.com>, Dan Connolly <connolly@w3.org>, "public-html@w3.org WG" <public-html@w3.org>
On May 26, 2007, at 3:48 PM, Preston L. Bannister wrote: > Perhaps not a great practice, but there is a certain amount of > sense in this usage. In HTML4 yes, since <style> could not be scoped. In HTML5, with the addition of scoped <style>, unscoped style no longer makes sense outside of the <head>. > If a page is dynamically assembled from multiple somewhat- > independent sources, then the need for dropping in new <style> at > the point the fragment is inserted into the page is understandable. Yes, but it should be scoped. > Not great for the browser implementors, but if in practice class > names are chosen so that the <style> only apply to the included > fragment, then effect should be limited in scope. > That's what scoping does. To reiterate, I am proposing that unscoped style outside of the head be non-conformant, but that scoped style is ok. Scoped style is a new feature of HTML5. http://www.whatwg.org/specs/web-apps/current-work/#the-style I am proposing that <style scoped> be allowed anywhere, but that <style> should only be conforming if used inside the head. I do not believe anyone has presented a valid use case for using unscoped <style> outside of the <head>, and given the performance impact it will have on the display of Web pages, the practice should be strongly discouraged. I think "scoped" should be even more limited than it already is, and that the HTML5 spec should limit the scoping to nodes that follow the <style> in a pre-order traversal of the tree. dave (hyatt@apple.com)
Received on Sunday, 27 May 2007 04:31:39 UTC