Differences

This shows you the differences between two versions of the page.

--- protected:template [2009/05/08 17:20] – created mark
+++ protected:template [2009/05/08 17:36] (current) – mark
@@ Line 2: / Line 2: @@
 ====== History ======
 ====== Algorithm Intent ======
+A description of the goals behind this
+software clustering algorithm.  Examples of such goals include:
+     * To decompose a software system into subsystems
+     * To recover the architecture of a software system
+     * To discover possible ways to restructure the software
+Explicitly stating algorithm intent can simplify selecting a software
+clustering algorithm for reverse engineers, as well as understanding a
+clustering algorithm for software clustering researchers.
 ====== Factbase Properties ======
+A description of the input that the software clustering algorithm expects (an algorithm's input is often called a factbase since it contains facts
+extracted from the software system).  For example, software clustering
+algorithms often expect that the factbase includes only dependencies
+between modules.
 ====== Clustering Objectives ======
+A description of the ideal output
+of the software clustering algorithm. Based on experience or heuristics,
+the designer of the algorithm decides what a meaningful decomposition of
+a software system should look like. For example, a software decomposition with maximum cohesion and minimum coupling.
 ====== Process Description ======
+A brief description of the main idea
+behind the algorithm's implementation. For example: "ACDC creates
+clusters by detecting established subsystem patterns in the given
+software system. Software entities that are not clustered this way are
+assigned to the subsystem that depends the most on them".
 ====== Decomposition Properties ======
+A description of the properties
+of the decompositions that the software clustering algorithm creates.
+These properties are either a direct result of the clustering objective
+(e.g. Bunch creates decompositions whose value of the objective function
+is higher than that for other algorithms) or an
+artifact of the algorithm's implementation (e.g. Bunch creates
+decompositions that are more balanced than most algorithms).
 ====== Algorithm Restrictions ======
+Since software clustering
+algorithms are heuristic algorithms, their performance
+may depend on the type of software system being clustered. Such dependencies
+need to be documented here. For example, Bunch may not be well-suited
+for event-driven systems.
 ====== Failed Assumptions ======
+A description of ideas that did not
+work very well while developing the algorithm. This information is
+often omitted in publications but could be invaluable to the software
+clustering researcher.
 ====== Detailed Algorithm Description ======
+ The nuts and bolts of the
+algorithm's implementation are presented here. This will often be the
+only section of significant length in an algorithm description.