HTML paste cleaner

The clipboard carries more than one version of what you copied. There is a plain text flavour, which is the words, and an HTML flavour, which is what an editor actually takes. Word's HTML flavour is built for round-tripping back into Word, not for the web, so it includes conditional comments wrapping whole stylesheets, an o: XML namespace, mso- style properties on almost every element, and MsoNormal classes throughout.

A sentence and a link can arrive as several kilobytes. It renders acceptably in the editor you paste into, which is why it survives, and then causes trouble later: Gmail clips messages over roughly 102KB, and the junk counts toward that.

Removed: Office conditional comments and the XML islands inside them, o: and w: tags, style and meta and link elements, xmlns attributes, Mso class names, mso- properties inside style attributes, font tags, and spans that carry nothing at all.

Kept: your text, your links, emphasis and structure, and any real CSS. A style attribute holding both a colour and an mso- property loses only the mso- part, because the colour is usually something you chose. Nothing is rewritten, so the cleaned HTML says what the original said with less around it.