Molecules and atoms

Markup only goes so far Battles rage periodically over how much markup is appropriate to any given document. James Tauber performed a nice reductio ad absurdum by suggesting that all content be marked up letter-by-<letter>l</letter><letter>e</letter><letter>t</letter><letter>t</letter><letter>e</letter><letter>r</letter>
Different perceptions of "normal" Database normalization created enough confusion - still does. Different developers have rather different perceptions of what qualifies as "normal" XML. My adequate markup is far too general for some, while other's adequate markup bristles with so many angle brackets that it looks downright spiky.
Your atom is my molecule ISO8601 dates look like atoms to some people, and the classic integer representations of date/time information certainly reduce complex information to a single very large number.
Special needs Some people really want to see documents marked up with every word and every sentence identified as elements. Doing that markup by hand is a huge headache, but doing it automatically isn't especially difficult in a large number of cases. (There are, of course, some extremely difficult cases.)