Review of NMF - Artur Boronat

[arturboronat]

Does the solution highlight a promising research direction?

• This solution aims at illustating advantages of model transformation w.r.t. conventional programming languages - focussing on incremental, bidirectional synchronization. This is a strong point in favour.
• The paper describes the motivation for the solution very well but does not evaluate the proposed advantages: times used for initial/synchronization phases, types of updates used, etc. This weakens the point above substantially.

Correctness

• NMeta requires a different format for metamodels/models and the proposed EMF-based validator cannot be used to check the correctness of the transformation. As far as I'm aware this has not been reimplemented for this solution.

Completeness

• The solution addresses the problem of mapping a truth table, representing a set of boolean functions, to a BDT. In the paper, clarify which metamodel is being used and that BDTs are produced.
• Synchronization works for new additions of assignments only and propagation of updates (changing the value of an assignment in a truth table that is already mapped) are not supported.

Understandability

• In the paper, the declarative part of the synchronization specification is reasonably straightforward to understand. However, I'm missing the outline of the transformation - the key parts of the transformation are explained along the imperative part.

Performance

• Run time is not logged.
• I could execute up to GeneratedI14O4Seed7634 in about 10-15 minutes.
• I'm assuming that the solution is being penalized by the initial synchronization that is performed for each input truth table and that subsequent propagation of updates will be much faster. Hence, without the discussion on incremental execution (the point main goal above), the opponent tends to think that the solution is not very efficient - happy to be convinced otherwise!

Optimality

• Output models could not be accessed to check results and to compare output model sizes.

Quality of software artifacts

• The solution could be executed without problems.
• There is no readme file explaining how to identify the different parts of the solution: input models, different components of the solution, etc.