The experiment.

Characters such as 'pound sign' are outside the Basic Latin set that is common to UTF-8 and ISO-8859-1 so this page is to test what happens if you try to represent them in various ways.

First, a UTF-8 character in the source XML - £. The HTML version of this page will have been run through XSLT with output to HTML/ISO-8859-1 so that character should have been converted before upload - but is it the ISO-8859-1 character code, is it a character reference? No, it seems to have been converted to an entity reference! In the XML version it should still be UTF-8 so does it get through safely in the XML version of the page?

Second, a character reference (£) in the source XML - £. In the HTML version the reference should have been handled by the XSLT before upload so the result should be as above. In the XML version, the character reference should be untouched and so handled by the browser's XML parser.

Attempting to use the entity reference £ should fail on both routes - £ - except that I put an entity declaration in the file. My local XSLT processor would otherwise complain when I try to generate HTML and the browser's XML parser should complain for the version served as XML (Firefox did when I tried it).