It is relatively easy to develop a prototype of a Korean language processing system, but it is very difficult to make it an operational system. In this paper, we survey the current status and methodological issues of the Korean language processing systems such as morphological analyzer, parser and machine translator. In most cases, Korean language processing system easily comes to a dead-end state where its performance can not be improved any more. The reason is that it adopts a general algorithm covering similar problems as a whole because specific low-level problems are not clearly defined and their algorithms are unclear. So, when we add some restrictions to solve an individual linguistic problem, they are also applied to other linguistic phenomena as a side effect. It causes a critical problem that the improvement of the algorithm is very difficult. This paper proposes a 2-step paradigm, a divide-and-conquer method by the functional modularization, a simplification method, and an exception handling technique to develop an operational system that does not fall into a dead-end state.