The distance between what they teach us at the university and what happens in the real world is sometimes very large. And I can say this from my personal experience as a lecturer in an Information Systems Analysis and Design course at the Technion. I’ve taught this course for 3 semesters (including this). The course teaches model based/driven/whatever analysis and design of software systems using UML and OPM. I didn’t choose the course, it chose me (it is the course my supervisor teaches it, him being the creator of OPM). We usually divide the course in two, he teaches OPM and I teach UML.

So for the last 3 semesters I have been teaching UML to undergraduate students. And the main problem is that before I was in the academia I worked 9 years as a software engineer (programmer/architect), and almost never found UML to be of any use (except for class diagrams which are really good) Never in my 9 years of work in a big software entity did I see someone create a state-machine, sequence diagram or activity diagram before coding. Nobody though it was useful, not even management.

How can I keep a straight face in front of the students? I can because I believe that there is a problem of lack of abstraction in software development and someone needs to fix it. But is UML modeling the solution? If UML is promoted by so many academics and companies as the solution to all of our problems, then maybe I am missing something. Maybe there is some proof out there that UML does improve software development, that UML is useful.

Therefore I embarked myself in the mission to find this proof by doing a “systematical literature review” of all works that have tested the “usefulness” of UML either by experimenting with mice (i.e. undergraduate students) or in the industry. I search the main publishers of software engineering articles (IEEE, ACM, Springer) since 2005 (year UML 2.0 was introduced), read 3000+ titles and probably half the amount of abstracts to filter these articles, reducing the number of possible matches to 223 articles. Now I am fast-reading the articles to see if they are really empirical or their title/abstract was misleading. So fat 70/114 did do some empirical evaluations, but most of them between “variations” of UML. But I still haven’t found what I’m looking for, which is real evidence that UML does improve software development (there are 2 industrial projects where the usage of UML was seen as beneficial, but that is NOT proof). And I am really skeptical that I will find any proof.

WOW! This post just went in a completely different direction from the one I wanted it to take… I started writing this because of something I read a couple of weeks ago about the distance that exists between the academia and the industry in software engineering. I guess my view is like theirs, and I wanted to add my personal touch.

Oh well, back to filtering UML articles