368437 - XMLSerializer should serialize HTML as XHTML (namespace, lowercase tags)

Status: RESOLVED FIXED

(Core :: XML, enhancement)

Description

[Jesse Ruderman]

Attachment: testcase

When reducing testcases that involve DOM manipulation, I often want to know whether a bug can be triggered by the final DOM, sans scripts.  To find out, I have to serialize the DOM and load it.  XMLSerializer.serializeToString does this very well, except when the page was initially text/html.

For text/html, XMLSerializer.serializeToString gives me uppercase tags and no namespace.  This isn't very useful because:

* If I feed it to the HTML parser, the HTML parser loses XUL/SVG/MathML elements and tries to be too clever when tags are nested strangely (e.g. block-in-inline or missing <tr>).  Also, <script src="..."></script> in the original HTML becomes <SCRIPT src="..."/>, which can cause the HTML parser to ignore the rest of the file.

* If I feed it to the XML parser, the tags are treated as being from an unknown namespace, and I get the "unstyled XML" display.  The uppercase tags prevent me from fixing this by simply adding an xmlns.

Steps to reproduce:
1. Load the testcase.
2. Click the button.

Result: <TABLE border="1" id="n"><TBODY><TR><TD>Foo</TD></TR></TBODY></TABLE>

Expected: <table xmlns="http://www.w3.org/1999/xhtml" border="1" id="n"><tbody><tr><td>Foo</td></tr></tbody></table>

Comment 1

[Jesse Ruderman]

pvnick and I wrote a script to serialize HTML as XHTML, so this isn't as important for me anymore.

Comment 2

[Jesse Ruderman]

Fixed by bug 468708, but bug 500937 will probably change the behavior again.