Communications of the ACM
Time to Resurrect PSP?
Let us assume for the sake of the argument that software quality matters. There are many ingredients to software quality, of which one must be the care that every programmer devotes to the job. The Personal Software Process, developed by Watts Humphrey in the 1990s [1], prescribes a discipline that software developers should apply to produce good software and improve their professional ability over their careers. It has enjoyed moderate success but was never a mass movement and rarely gets mentioned nowadays; few software developers, in my experience, even know the name. Those who do often think of it as passé, a touching memory from the era of Monica Lewinsky and the Roseanne show.
Once cleaned of a few obsolete elements, PSP deserves to be known and applied.
PSP came out of Watts Humphrey's earlier work on the Capability Maturity Model (see my earlier article on this blog, What is Wrong with CMMI), a collection of recommended practices and assessment criteria for software processes, originally developed in the mid-eighties for the U.S. military contractor community but soon thereafter embraced by software outsourcing companies (initially, Indian ones) and later by other industries. Responding to complaints that CMM/CMMI, focused on processes in large companies, ignored the needs of smaller ones, and lacked individual guidance for developers, Humphrey developed TSP, the Team Software Process, and PSP.
The most visible part of PSP is a six-step process pictured in the middle of this diagram:
The most visible and also the most corny. Who today wants to promise always to follow such a strict sequence of steps? Always to write the code for a module in full before compiling it? (Notice there is no backward arrow, the process is sequential.) Always to test at the end only? Come on. This is the third decade of the 21st century.
Today we compile as we code, using the development environment (IDE) as a brilliant tool to check everything we do or plan to do. For my part, whenever I am writing code and have not compiled my current draft for more than a few minutes I start feeling like an addict in need of a fix; my fix is the Compile button of EiffelStudio. At some eventual stage the compiler becomes a tool to generate excutable code, but long before that it has been my friend, coach, mentor, and doppelgänger, helping me get things (types, null references, inheritance...) right and gently chiding me when I wander off the rails.
As to tests, even if you do not buy into the full dogma of Test-Driven Development (I don't), they get written and exercised right from the start, as you are writing the code, not afterwards. Compile all the time, test all the time.
It's not just that a process such as the above ignores the contributions of agile methods, which are largely posterior to PSP. As analyzed in [2], agile is a curious mix of good ideas and a few horrendous ones. But among its durable contributions is the realization that development must be incremental, not a strict succession of separate activities.
This old-style flavor or PSP is probably the reason why it has fallen out of favor. But (like agile rejection of upfront lifecycle activities) such a reaction is a case of criticism gone too far, ignoring the truly beneficial contributions. Ignore PSP's outmoded sequence of activities and you will find that PSP's core message is as relevant today as it ever was. That message is: we should learn from the practices of traditional engineers and apply a strict professional discipline. For example:
- Keep a log of all activities. (See "Logs" in the above figure.) Engineers are taught to record everything they do; many programmers don't bother. This practice, however, is essential to self-improvement.
- Keep measurements of everything you do. (There are lots of things to measure, from hours spent on every kind of task to bugs found, time to fix them etc.)
- Estimate and plan your work.
- Clearly define commitments, and meet them.
- Resist pressure to make unreasonable commitments (something that agilists approach also emphasize).
- Understand your current performance.
- Understand your programming style and how it affects various measures. (As an example, code size, as a function of the number of routines, depends on whether you are more concise or more verbose in style).
- Continually improve your expertise as a professional.
PSP does not limit itself to such exhortations but gives concrete tools to apply the principles, with a view to: measuring, tracking and analyzing your work; learning from your performance variations; and incorporating the lessons learned into your professional practices. On the topic of measurement, for example, PSP includes precise guidelines on what to measure and how to measure it, and how to rely on proxies for quantities that are hard to assess directly. On this last point, PSP includes PROBE (PROxy-Based Estimating, you cannot have a method coming out of the world of US government organizations without cringeworthy acronyms), a general framework for estimating size and resource parameters from directly measurable proxies.
This is what PSP is about: a discipline of personal productivity and growth, emphasizing personal discipline, tracking and constant improvement. It is not hard to learn; a technical report by Humphrey available online [3] provides a sufficient basis to understand the concepts and start a process of self-improvement.
Watts Humphrey himself, as all who had the privilege to meet him can testify, was a model of seriousness and professionalism, the quintessential engineer. (I also remember him as the author of what may be the best pun I ever heard -- ask me sometime.) PSP directly reflects these qualities and -- ignoring its visible but in the end unimportant remnants from outdated technical choices -- should be part of every software engineering curriculum and every software engineer's collection of fundamental practices.
References
[1] Watts Humphrey, Introduction to the Personal Software Process, Addison-Wesley, 1996.
[2] Bertrand Meyer: Agile! The Good, the Hype and the Ugly, Springer, 2014, see here.
[3] Watts Humphrey, The Personal Software Process, Software Engineering Institute Technical Report CMU/SEI-2000-TR-022, available (in PDF, free) here.
Bertrand Meyer is chief technology officer of Eiffel Software (Goleta, CA), professor and provost at the Schaffhausen Institute of Technology (Switzerland), and head of the software engineering lab at Innopolis University (Russia).
Comments
Andres Rodriguez Guapacha
Learning PSP is a challenge for me. Back when I was in college, the materials were not easily available and instructors were even more difficult to find. Ten years later and SEI doesn't even bother to mention PSP anywhere in their homepage unless you specifically search for it. But even when I could get the book, I found it a bit hard to read and difficult to use with some agile practices that I think are also worth it, like you mention in your article. I might be fortunate because I just found some new content that seems to address this last issue: SEI added some content in 2020 called PSP for the 21st century. ( https://resources.sei.cmu.edu/library/asset-view.cfm?assetID=540135 ). I will give it a go and restart my PSP self-learning.
Displaying 1 comment