home > plop > 2006 > Papers > Notes > 5-notes.php

PLoP 2006 Shepherded Papers

Notes on Paper 5: REJECTED

I do NOT recommend accepting the research paper "Latent Binding of Pattern Specialization" to be presented/workshoped at the conference. This summary is based on the version submitted on July 30th.

Summary of the paper:

The paper claims that design patterns only handle the reuse of design without any concern for reuse of code. Therefore, every time when a developer wants to use a design pattern, she must pass through the "tedious" hurdle of implementing the design pattern into her codebase. As a solution, the authors propose a separation of concerns between patterns and their particular binding to a specific code base. In the design phase, designers can use some skeletons of patterns, without referring to the concrete classes that will eventually be the contributors of the patterns. In the implementation phase, developers can reuse those pattern skeletons (provided as a library of reusable pattern skeletons) and bind the concrete codebase classes to these design skeletons.

Problems/Flaws:

Besides the writing problems, there is something at the conceptual level that I could not agree with. First, I have used patterns many times as a developer. It never felt to me that using patterns was a burden because I had to apply them to my particular context. I had a hard time explaining the authors that patterns are not about code reuse, but about reuse of ideas. If they are looking for both kinds of reuse, frameworks provide reuse of both code and ideas. Second, any solution that is not considerably simpler or requires considerable less time than implementing the "classic" design pattern, is not worth using.

Improvement during shepherding:

The paper originally started as a "pattern". I strongly suggested that they drop the misnomer "pattern" (since they did not present a pattern) and instead write a research paper on patterns. The authors followed my advice. However, there are many other times during the three iterations that we went through, that the authors seemed to pay no attention to my lengthy and detailed suggestions for improvement. I felt many times that I literally wasted valuable hours, since the authors had a hard time incorporating my feedback. After explicitly complaining myself to the authors, they seem to have incorporated some of my advice on the latest version (July 30th).

Relevance to the research literature on patterns:

The writing is weak compared to the standards of a research paper:
- Introduction. The paper is poorly motivated. The reader is not convinced that the problem they try to solve is real/important.
- Related Work. Although the paper cites 12 related work, the treatment of the related work is poor too. The related work should map the research territory for a new reader in this research field. This paper fails to do so.
- Evaluation. There is not a hint about how would one evaluate their claim that the proposed solution could reduce the complexity and time required to implement design patterns. In my opinion, the extra work that a developer needs to bind the concrete classes to the library pattern skeletons is no simpler or no quicker than implementing the classical pattern. It looks like the authors are not even concerned with evaluation.
- Research contribution. I could not see anything that is researchy/hard in their proposed solution. This paper leaves a hard-core researcher asking himself why did he wasted time reading the material. <<<<<<<<<<<

best,
Danny