Re: Input on the agenda

On Wed, Mar 18, 2009 at 4:09 AM, Simon Pieters <simonp@opera.com> wrote:
> On Wed, 18 Mar 2009 11:21:01 +0100, Jonas Sicking <jonas@sicking.cc> wrote:
>
>> Note that we wouldn't have to "break" style fully. We could still give it
>> the same treatment as <script>. That means that inline scripts with
>> <![CDATA[]]> would work fine.
>
> ...if they used //<![CDATA[ and //]]> (or /*<![CDATA[*/ and /*]]>*/)? Or do
> you want to make the JS engine and CSS parser aware of these strings (like
> <!--)? Or do you want the HTML parser to strip these strings inside CDATA
> elements?
>
> Personally I think it would be ok to require authors to escape these strings
> in text/html, just like they have to do for XHTML as text/html today.

The problem is that that that doesn't support the use-case of copying
existing XML-SVG into text/html, since no such SVG uses //<![CDATA[

I see two solutions:

1. Support <![CDATA[]]> while tokenizing CDATA.
2. After having read the data from the DOM, strip a leading
'<![CDATA[' and trailing ']]>' before handing the data to the CSS/JS
engine.

>> As would any stylesheets with no entities. One thing that would break is
>> stylesheets without <![CDATA[]]> that use the child selector ( > ).
>
> (Only if it was escaped as &gt; which is not required in XML.)

Good point, that would even further reduce the risk that pages out
there would break.

/ Jonas

Received on Wednesday, 18 March 2009 17:33:28 UTC