++ thank you. That seems like a nippy way of doing it (compared to creating a HTML parse tree, anyway). It might also help me get over my phobia of the /x regex modifier.

However, I'm going to be an XHMTL pedant and point that there's a few things it doesn't handle correctly. By correctly, I mean, the end result isn't identical, from an XML parser's point of view, with the start.

• It needs to leave CDATA sections alone. In XHTML Strict, SCRIPT and STYLE sections may declare the content to be unparsed character data. This is useful because it allows you to have '<' and '>' in your scripts (eg the Javascript comparison operators) and styles (CSS contextual selectors) without having to escape them.
• It shouldn't touch the whitespace at all within PRE elements; inside these, whitespace should be taken as literally in the file, and not closed up. For example, this will come out wrong:

<pre>    <span id="foo">foo</span>    </pre>
[download]

Sorry, I don't want to detract from a really nice piece of work; I can see that it would definitely be useful in more data-oriented XML settings. However it's not really accurate enough for me to use in a production setting.

cheers
ViceRaid