Yeah, that's the other thing. Invent your own minimalist encoding. Its gross and yucky and you might prefer to use a smaller character set like latin1 instead of utf-8. You'd at least be swapping all the utf-8 rules for validity with something that stays with the octet / character limit. Heck, why not encode for ASCII? You'll be specifying your XML file's encoding in the preamble so there isn't any ambiguity about the data representation on the consumption side. Its seven bits and nice to look at.