UML: TIP: avoid revealing operations and attributes of Classes (and Interfaces, if possible) in more than one diagram (UNLESS you do so to illustrate an explicit collaboration with another Class with revealed Features) !

The reason is subtle yet important. If you add/remove/change an Operation or attribute Property you'll have to manage that change in more that one diagram. It is easy to forget that it appears in more than one place, and you'll find diagrams you spent time perfecting suddenly graphically corrupted. This can happen, for example, on reverse-engineering changing code.

The principle can be relaxed a little for Interfaces, the operations of which might need communicating in more than one diagram, however for Classes one does well to apply it as strictly as possible.
Exception: when eliciting analysis Classes during domain analysis one can make a case for revealing and collecting Operations and attribute Properties in one place in large overview diagrams.
Related notes