914545 - "undefined entity" error triggered from JavaScript in XHTML page

Status: RESOLVED INVALID

(Core :: DOM: Core & HTML, defect)

Description

[Vladimir Pycha]

User Agent: Mozilla/5.0 (X11; Linux i686; rv:23.0) Gecko/20100101 Firefox/23.0 (Beta/Release)
Build ID: 20130814063812

Steps to reproduce:

I had a page served with "application/xhtml+xml" MIME type, with XHTML 1.0 Transitional doctype with an FPI (Formal Public Identifier).


Actual results:

XHTML entities normally worked, except that I was getting a pair of runtime JavaScript errors whenever I was assigning to either of properties element.innerHTML or element.outerHTML a string containing containing an XHTML entity, in particular " " entity. The script was inside a CDATA section.

The pair of JavaScript errors were:

undefined entity
SyntaxError: An invalid or illegal string was specified

After I removed the entity from the string, I was getting neither of these errors.


Expected results:

No JavaScript errors should have occurred.

Comment 1

[Vladimir Pycha]

I created a test case, see the URL above.

Now I tested it in Google Chrome and Internet Explorer 9 and the results were the same, the same errors occurred.

So, which RFC dictates that browsers should behave this way that does not make sense?

Do I really have to be using " " entity in my scripts instead of " " and have to tell Google to do so too in their Custom Search scripts?

Comment 2

[Vladimir Pycha]

A related question and answer at stackoverflow:

http://stackoverflow.com/questions/4162270/dom-exception-when-assigning-html-entities-to-innerhtml

Comment 3

[Boris Zbarsky [:bzbarsky]]

> So, which RFC dictates that browsers should behave this way

The relevant spec is http://domparsing.spec.whatwg.org/#extensions-to-the-element-interface which calls http://domparsing.spec.whatwg.org/#concept-parse-fragment which calls http://www.whatwg.org/specs/web-apps/current-work/multipage/the-xhtml-syntax.html#xml-fragment-parsing-algorithm

It looks like we're correctly doing what this says: it never says to import the context element's document's DTD into the new XML parser.  In fact the spec clearly says this is purposeful:

  Note: If there is a context element, no DOCTYPE is passed to the parser, and
  therefore no external subset is referenced, and therefore no entities will be
  recognized.

If you think the spec here should change, please raise a spec issue...