I always have troubles with UTF8.

The scenario: Mulitple sources (like xml, csv, own cms), and one database: Mysql. Information needs to be displayed on a website or outputted in XML to another program.

Do I need to store text like they are? Special settings needed (collocation?) in the mysql database, what php functions need to be used?
To display these on a webpage I use a utf-8 header using php or a meta tag saying it is utf8. But still special characters are displayed incorrect or the xml is broken because of an invalid character.

What is the best consistent way to avoid these problems, or has someone really good reading about this topic?