A frequently found model of a scientific paper is the following. The work describes a new idea, prototyped perhaps in a small system. The claim to ``scientificness'' is then made by feature comparison. The reports sets out a list of features and qualitatively compares older approaches with the new one, feature by feature.
I find this method satisfactory when a radically new idea or a significant breakthrough is presented, such as the first compiler for a block-structured language, the first timesharing system, the first object-oriented language, the first web browser. Unfortunately, the majority of papers published take much smaller steps forward. As computer science becomes a harder science, mere discussions of advantages and disadvantages or long feature comparisons will no longer be sufficient. Any PC magazine can provide those. A science, on the other hand, cannot live off such weak phenomenological inferences in the long run. Instead, scientists should create models, formulate hypotheses, and test them using experiments.