When I suggested a system of codes to enter into coding boxes, I made reference to a modified version of the Library of Congress Classification system. This is not to suggest that it is any better than anything else; it's just an established system that is a convenient jumping off point. Starting from scratch would involve reinventing a lot of basics. The LCC is a system based on one, two and sometimes three letters, a number up to four digits long, a decimal point, and decimalized alphanumerics of varying lengths. For practical purposes I see no reason to go beyond three letters at the beginning, if at all. The only conceivable exception for the future would be for a very big subject area that is just crying for further subdivision. We would not be using the system to put numbers on the spines of books to ensure that the books are put on the right shelf in the library.
The classification system should allow nesting of categories. Under the existing LCC "Q" represents sciences in general, "QA" represents mathematics and no three letter codes are defined in the QAs. Without prejudice, this would leave free to define "QAG" to represent geometry. From the searcher's point of view, he could find his geometric subject by searching either Q or QA or QAG, but the result from searching on Q could be big and mostly useless to him. An article could be classified in more than one category; one that deals with both calculus and geometry could show both a QAG and a QAC category. Also, a person who wants to be able to contribute through classifying articles should have the option to choose only those items which are unextended when he's looking for work.
Ib my previous post I mentioned code "AAA" for unclassified, but there are also other codes that could serve Wikipedia's own special purposes.
Eclecticology