When adding a p element with a text node containing an html code, the text shown on the page is the litteral string, without any interpretation of this html code (see attachment for test reduction)
Created attachment 22086 [details] Test case
createTextNode is not expected to interpret entities. Entities are a function of the HTML and XML serializations of the DOM, not of the DOM itself. You can insert non-ASCII characters by passing a string to createTextNode that contains the required Unicode characters. If you need to insert an entity reference then the DOM way of doing that would be to use Document.createEntityReference. Note that in my testing, WebKit behaves identically to Firefox 3 on this matter.