- From: Anne van Kesteren <annevk@opera.com>
- Date: Tue, 24 Mar 2009 17:20:51 +0100
- To: www-style@w3.org
Back in 2005 IRIs were discussed on this list and I did not quite get what Björn meant back then due to not understanding Unicode Normalization very well: http://lists.w3.org/Archives/Public/www-style/2005Mar/0096.html Anyway, four years forward and URL handling in CSS is still undefined. I did some testing today and CSS handling of URLs appears identical to HTML handling of URLs in browsers, except that the query component is always encoded in UTF-8 and not dependent on the encoding of the CSS document (or HTML document for that matter). (How browsers work follows the "convential wisdom" Björn describes rather than RFC 3987, by the way.) HTML URL handling is currently detailed by HTML5 but will likely be split out by Dan Connolly into a separate draft. I suppose it's best to wait and reference that draft when it is ready. We also need to make some textual changes of course to say what happens when the URL cannot be resolved, that the URL character encoding is always UTF-8, etc. The difference from referencing the IRI specification is that 1) URLs found in non-Unicode encoded documents are not normalized to NFC and 2) it actually deals with URL strings such as url("image.php?name=CAT HAZ BURGER") (i.e. that contain illegal characters). (There are some small interop issues. IE6 (not sure about newer versions) sends the query string as raw UTF-8 bytes rather than having them percent-escaped. WebKit normalizes text encoded for sending out to NFC, which includes url() in CSS as it happens.) -- Anne van Kesteren http://annevankesteren.nl/
Received on Tuesday, 24 March 2009 16:21:38 UTC