Java Runtime Libraries Comparison

General

Main Concept: The main idea is to compare the three most complete implementation of the 1.6 Java Platform Specification JSR. The libraries will be compared in the terms of software complexity metrics (CK suite and some others). Possible publication venue: ICSE.

Participants: OpenJDK, GNU Classpath, Jakarta Harmony.

Brainstorming

Various ideas that can be evolved in the process of this paper.

Open Source Vs Closed Source Development

Nice and proven, the problem with that, it is done before in various contexts.

Mathematical Approach (Vs Public Interface)

Maybe we can find (observing our data ofc) a mathematical equation or a theorem of some sort, that will describe certain relations between the metrics.

External Characteristics

• Micro-benchmarking (maybe in collections etc). (update: maybe this should be a separate paper)
• Number of Bugs (update: cannot find data for OpenJDK)

Requirements Drives Quality

Examine if Requirements dictates quality or the design of code, since in our case we have a specification that guides the implementation.

• Clustering by public class + non-public dependencies

Threats of Validity

• The three systems are not directly comparable (updated: added in the paper)
  • OpenJDK implements 1.6 JSR,
  • Jakarta Harmony implements 1.6 JSR (beta)
  • GNU Classpath implements part of 1.5 so far (updated 9/9/2008)