Large Scale Empirical Validation of the Aspect-Oriented Design Hypothesis

点击翻译按钮获取中文摘要

Throughout the history of technology, modular design has proven to be an effective way to deal with system complexity. Modular design is also routinely applied to software-intensive systems. Software, however, is different from hardware. Specifically, the fact that software can programmatically affect other software broadens the options for modularization. The very nature of modules can be redefined in ways that would be impractical in hardware. Aspect-Oriented Programming (AOP) has recently put forth one of those redefinitions. In AOP, some modules called ''aspects'' directly address the crosscutting nature of some design concerns by modeling those concerns from outside the modules to which the local effects belong.This research will conduct a large-scale empirical validation of the design hypothesis put forth by AOP, and will leverage it to derive principles for modular design. The AOP hypothesis has three sub-hypotheses: (1) Complex software must cope with the existence of cross-cutting concerns; using traditional procedure- or object- oriented design, those cross-cutting concerns show up as design elements scattered throughout several modules and tangled with other concerns within those modules; (2) Scattering and tangling are ''bad'' for the design process; (3) The alternative composition mechanisms embodied in an aspect-oriented language like AspectJ are ''''better'''' than the traditional ones. The empirical validation of AOP will be enabled by an infrastructure called Sourcerer, developed for collecting, searching, and analyzing software on the very-large scale of Open Source software available on the Internet. More broadly, this research will investigate methodologies for conducting empirical research in Open Source software.