Assume L is the computation you would like performed by ConsoleAppender regarding the charset. I am saying that L can be performed on the fly at configuration time (that is while reading logback.xml) and the result of this computation assigned to a variable named according to your preferences, say CONSOLE_CHARSET. You would than inject CONSOLE_CHARSET as the charset to be used by LayoutWrappingEncoder within ConsoleAppender.

Does this make better sense?