An earlier work on automated optimization heuristics for cryptanalysis of classical ciphers proposed a few algorithms for that task (e.g. genetic, simulated annealing, tabu search). A Java-language open-source Cryptolysis project has implemented these algorithms for verification and comparison purposes in a consistent frameworked environment allowing for additional algorithms. Another Java-language open-source project, MARF, has collected a number of frameworked classification algorithms (e.g. distance, neural network, similarity measure, etc.). We extend Cryptolysis with the wrappers for the algorithms implemented in MARF to add to the heuristics collection new results and compare them with the previously implemented algorithms. Conversely, we improve MARF's implementation by porting the Cryptolysis's implementations of search algorithms for various classification tasks in natural language and others. Additionally, we improve the system with the natural language word segmentation for the deciphered text corpora that lacks spacing and punctuation. As a result this work we validate the software architecture and design used in both frameworks as a sustainable and correct approach and practice that should be followed and re-enforced for such frameworks.