To: Alan Kay
From: The Program Committee
Dear Dr. Kay,
The program committee thanks you for the submission of your paper “Object Orientation: A New Paradigm of Programming”. Unfortunately your paper has been rejected. We had many fine submissions this year, but as you know we must accept no more than 15% of submissions to be considered a premier conference. The reviewers’ comments are attached below.
Reviewer 1 comments: You propose three new language features: encapsulation, polymorphism, and inheritance. Even though your paper was the maximum 12 pages, it discussed each of these concepts only informally, and did not do any rigorous evaluation. The general rule is that evaluation should comprise 75% of the paper, leaving 3 pages to explain your ideas. To do proper Computer Science, you must evaluate new ideas incrementally. I recommend you write three separate papers each evaluating the addition of one of these features to Fortran and doing detailed case studies of the result. In the future, please avoid excessive novelty.
Reviewer 1 evaluation: FAIL
Reviewer 2 comments: You claim that object orientation is in some sense more “natural” and “intuitive” than procedural programming, but offer only anecdotes and hand-picked examples as justification. To do proper Computer Science, such claims must be empirically evaluated. I suggest you do user studies on experienced Fortran and Cobol programmers, measuring how well they can write scientific and accounting programs in your new language compared to existing languages. I strongly suspect you will get negative results, because such professional programmers will find your ideas not at all natural, but in fact quite different from the proven techniques of programming they have mastered. I will not even bother to criticize the claim of having invented a new paradigm, for I can think of no experiment to falsify that statement, rendering it scientifically meaningless.
Reviewer 2 evaluation: FAIL
Reviewer 3 comments: You propose to add layers of overhead on every subroutine call, and worse, to dynamically allocate working storage. Yet you offer no controlled performance benchmarks to show that you can compete with proven technology. Memory and CPU are scarce resources, and will continue to be so for the foreseeable future. In my opinion, the performance of the language features you propose could only become competitive after 20 years of further research. Such long time horizons are out of the question, both for those in industry as well as those on a tenure track. Proper Computer Science demonstrates immediate quantifiable benefits.
Reviewer 3 evaluation: FAIL
Reviewer 4 comments: The paper shows an ignorance of important related work. For example, there is extensive prior research on inheritance in Legal Software. Also, polymorphism reminds me of polynomials, the topic of my dissertation, which wasn’t cited. When I summarized your paper in 5 minutes to the committee members who hadn’t read it, one of them said it sounded just like what Joe Blow did in the 80’s, so it appears your work is neither novel nor properly researched. To do proper Computer Science you must demonstrate knowledge of the research literature and show how your work is a contribution in addition to it, even if it only seems vaguely similar after reading the paper’s title.
Reviewer 4 evaluation: FAIL