-Formal Model Driven Engineering (MDE) can be considered incongruent with Agile methodologies. However, with the advent of Agile, Software Engineering educators have an obligation to teach Agile development. Many instructors do so by employing experiential learning through Agile classrooms and projects. Teaching formal MDE and convincing students of its benefits can be challenging in such environments. In this paper, we discuss this position by considering established best practices in modeling education and their compatibility in Agile classrooms/projects. We argue that more than half the practices present some challenge in Agile environments, for which we provide initial suggestions. The most significant challenges are those pertaining to the MDE education practices of tailoring the MDE processes to participant knowledge and course context, defining the types of MDE artifacts before the project or course begins, and ensuring industrial relevance to prepare students for post-graduation. Discussion of these challenges at the symposium provided additional insights and suggestions, which we summarize in this paper. Notable suggestions include tailoring both the MDE and Agile processes to complement each other, employing a minimal set of MDE artifacts necessary for code generation, and having a large initial Agile iteration to form a realistic minimal viable product, respectively. We plan on applying these MDE educational best practices in our Agile course offerings, incorporating our initial suggestions and those from the symposium. Our goal is to stimulate discussion on these challenges and others, and help guide future research and course offerings.
Agile development methodologies are widespread in
Software Engineering. 35% to 45% of software developers use
some form of Agile, which is more than the 30.6% that
use nothing, 21% that use iterative, and the 13% that use
waterfall [
One strength of Model Driven Engineering (MDE) is code
generation [
This paper describes our position that following established practices for teaching MDE in an Agile setting presents unique considerations and challenges that should be addressed by the modeling education community. Our discussion considers these practices in the context of Agile, and includes suggestions when conflicts exist. We include our own suggestions and those that arose at the symposium. Our goal is to stimulate ideas on these challenges, generate others, and guide future research and practice.
The paper begins in Section II with a quick primer on Agile Methodologies and Education, and MDE education. In Section III we argue Agile Modeling is inadequate for educating students in both Agile and Modeling. Our discussion begins in Section IV, which frames the conversation in terms of eight established formal MDE education practices. We conclude the paper in Section V.
Agile methodologies involve principles, practices, and
methods for software development and other fields. It allows teams
to be more flexible, and capable of meeting the demanding
needs and changing requirements of consumers. While not
applicable to all projects and products, it has seen notable
adoption recently [
There are different approaches to teaching students Agile
methodologies including using a decision support process with
use cases [
The importance and impact of teaching MDE has been
discussed by Hamou et al. [
One of the main considerations Hamou et al. identify when
teaching MDE involve the “multitude and the complexity of
the (MDE) concepts” [
P2 Defining artifacts and creation procedures beforehand.
That is, will students be using use case diagrams, state diagrams, sequence diagrams, et cetera.
P3 Consistency awareness and management to ensure all artifacts are consistent with one another. This can include establishing consistency rules for the students to use throughout the semester.
P4 Teaching the importance and usage of models and their elements, including relation of elements to code generation. A key principle to impart is that ”if the model is not used anywhere, it should not be created.” P5 Receiving and encouraging constant feedback from students. The course must be at a level students can appreciate and follow. Eliciting feedback from students is helpful in realizing this.
P6 Industrial relevance, that is, preparing students for the “real” world. The problems covered in the course and in projects should be real-world problems.
P7 Performing research and academic experiments. Kuzniarz and Staron provide an example of their experiments that assessed the usability of UML stereotypes to help increase UML model comprehension.
P8 Including new research and industrial trends in the course material and project. The purpose is to raise the students’ interest in emerging technologies and developments in MDE. Also, to have them better prepared when they graduate and pursue either industrial or academic pursuits. We consider these practices and their plausibility in Agile courses/projects.
Agile modeling is a light-weight modeling technique
intended to derive some benefits from modeling in Agile
settings [
The modeling community and Software Engineering
educators should teach students formal MDE to best prepare them
to work on safety critical, secure, and reliable software [
In this section, we discuss how Agile courses clash or fit with formal/generative MDE education practices. We frame our conversation in terms of the modeling education best practices mentioned in Section II-C, with each item, AX, corresponding to each PX in that list. For those that present challenges, we include suggestions based on our initial thoughts and our discussions at the symposium. This is only a starting point. We anticipate more challenges and considerations will emerge in future work.
This practice dictates instructors tailor development processes to a course and its context. However, Agile purists /proponents argue that a project process is either Agile or it is not; it cannot be “somewhat” Agile. While tailoring MDE aspects is possible, tailoring certain process-specific aspects may not be appropriate in an Agile setting. Thus, Agile courses and projects should tailor only MDE process aspects not Agile process aspects.
At the symposium, a participant argued MDE purists would
say MDE processes should not be tailored, and there are
several Agile methods that are highly customizable for
different contexts. A specific example of this is the Scaled Agile
Framework [
Defining MDE artifacts types for students before an Agile course begins is challenging. To teach MDE properly the choice of artifacts should be sufficient to facilitate and elucidate the formal MDE pipeline. However, the choice of artifacts must also be a feasible set students can create/update in each Agile iteration. More research and experimentation is required to address this.
The symposium attendees agreed that it is challenging to find the right balance of artifacts that are sufficient and feasible. One person posed the question if MDE is ever complete. Another person responded it should be capable of code generation at least.
Artifact consistency is easily possible in an Agile course or project. To enforce this, recurring Agile stories/tasks to manage and synchronize artifacts can be created automatically each iteration through tooling, like Jira1 or Trello2.
An industrial symposium attendee confirmed that they see this in industry. That is, they have explicate stories and tasks intended to ensure consistency. They verify correctness at the modeling level of abstraction.
Since each iteration in Agile requires a quality working product, teaching the effective use of MDE models fits nicely. However, considerations include, prior to the first iteration, 1) students should be familiar with MDE basics, and 2) code generation is setup and ready for them. This will likely require a primer on MDE and code generation before any work begins.
At the symposium, participants were in strong agreement that educators must teach, or review, ”good design” and modeling to students beforehand.
Adaption of the constant feedback best MDE education practice within Agile is seamless as this is a key Agile principle. Students will be in constant communication with the instructor/customer through both stand-ups and progress story boards.
The symposium attendees believe that constant communication both early on during the project and course will not impact MDE education, and having Agile-type retrospectives after milestones will be beneficial. No conflicts were anticipated.
Industrial relevance in teaching MDE applied in an Agile setting is challenging. MDE in formal domains, like automotive, telephony, and others is often front loaded. This is appropriate as requirements change less, and can be regulation/domain-driven. While students can learn the generative aspects in an Agile setting, it may not be as front loaded because Agile involves building small products and incrementally adding features. One way of tackling this may be having a large minimum viable product (MVP) in the first iteration.
1https://www.atlassian.com/software/jira 2https://trello.com
One of the symposium attendees from industry pointed out
they use their models mostly for diligence and auditing. So,
they were less concerned with students being familiar with
code generation. They develop their models in an Agile fashion
but not the product itself. Of course, this is organization
specific. Hutchinson et al. [
MDE Experimentation is possible in Agile courses/projects. For example, in class-wide projects, groups can swap models to analyze a model set’s ability to aid in comprehension each iteration. Or, in between iterations, groups can be asked to migrate their models to another tool to compare, teach, and experiment with multiple modeling tools.
There should be no problem in having students learn and employ emerging MDE research and tools in an Agile setting.
We summarize our qualitative assessment of these challenges in Table I. These helped form the basis for discussions at the symposium and, hopefully, beyond in generating more challenges and suggestions. Additionally, we plan on applying these MDE practices as best possible in our Agile course offerings in the near future for the purposes of observation and research.
Agile methodologies can be considered incongruent with formal MDE. Similarly, employing Agile education techniques can conflict with formal MDE education. As a starting point for discussion, we considered eight established best practices for teaching modeling with respect to their applicability in Agile classrooms and course projects. We postulate that more than half of these MDE practices face at least mild challenges in Agile settings. Some of our preliminary suggestions include automatic Agile task creation for artifact consistency, a larger first iteration release for industrial relevance, and others. We discussed these challenges at the symposium, and are conducting research by applying these practices and suggestions in our Agile class offerings in the near future. It is our hope that this paper helps guide future research and furthers the conversation on the educational interplay between Agile and MDE.
We would like to thank the Educators Symposium attendees for participating in our discussions, providing feedback, and brainstorming suggestions. We also want to express our gratitude to the symposium organizers for helping us facilitate the discussion.