Sorry for the late reply, but the point is that there is no trivial way to detect whether and how many times something has been encoded. You may end up with multiple levels of encoding in multiple systems and everything becomes untractable. Morever, as i said this doesn't have to be a problem, as you can just decode everything as much as you can BEFORE you put it in the db, as the db can handle all of that by itself. Just let it do its job. Paradoxically, if you use only channels that support utf8 and don't apply any transformation, your data is already perfect as it is. Then it is the job of the client to do what it needs to be able to render properly, but for instance a non-html client shouldn't need to use html libraries to be able to strip html stuff from the text before it can be displayed.