Posts Tagged ‘Watts Humphrey’

No Man’s Land

April 28, 2011 4 comments

Having recently just read my umpteenth Watts Humphrey book on his PSP/TSP (Personal Software Process/Team Software Process) methodology, I find myself struggling, yet again, to reconcile his right wing thinking with the left wing “Agile” software methodologies that have sprouted up over the last 10 years as a backlash against waterfall-like methodologies. This diagram models the situational tension in my hollow head:

It’s not hard to deduce that prescriptive, plan-centric processes are favored by managers (at least those who understand the business of software) and demonstrative, code-centric processes are favored by developers. Duh!

Advocates of both the right and the left have documented ample evidence that their approach is successful, but neither side (unsurprisingly) willingly publicizes their failures much. When the subject of failure is surfaced, both sides always attribute messes to bad “implementations” – which of course is true. IMHO, given a crack team of developers, project managers, testers, and financial sponsors, ANY disciplined methodology can be successful – those on the left, those on the right, or those toiling in NO MAN’s LAND.

l’ve been the anointed software “lead” on two, < 10 person, software teams in the past. Both as a lead and an induhvidual contributor, the approach I’ve always intuitively taken toward software development can be classified as falling into “no man’s land“. It’s basically an informal, but well-known, Brooksian, architecture-centric, strategy that I’d characterize as slightly right-leaning:

As far as I know, there’s no funky, consensus-backed, label like “semi-agile” or “lean planning” to capture the essence of architecture-centric development. There certainly aren’t any “certification” training courses or famous promoters of the approach. This book, which I discovered and read after I’d been designing/developing in this mundane way for years, sort of covers the process that I repeatedly use.

In my tortured mind (and you definitely don’t want to go there!), architecture-centricity simply means “centered on high level blueprints“. Early on, before the horses are let out of the barn and massive, fragmented, project effort starts churning under continuous pressure from management for “status“, a frantic iterative/sketching/bounding/synthesis activity takes place. With a visible “rev X” architecture in hand (one that enumerates the structural elements, their connectivity, and the macro behavior that the system must manifest) for guidance, people can then be assigned to the sparsely defined, but bounded, system elements so that they can create reasonable “Rev 0” estimates and plans. The keys are that only one or two or three people record the lay of the land. Subsequently, the element assignees produce their own “Rev 0” estimates – prior to igniting the frenetic project activity that is sure to follow.

In a nutshell, what I just described is the front-end of the architecture-centric approach as I practice it; either overtly or covertly. The subsequent construction activities that take place after a reasonably solid, lightweight, “rev X”, architecture (or equivalent design artifact for smaller scale projects) has been recorded and disseminated are just details. Of course, I’m just joking in that last sentence, but unless the macro front end is secured and repeatedly used as the “go to bible” shortly after T==start, all is lost – regardless of the micro-detailed practices (TDD, automated unit tests, continuous integration, continuous delivery, yada yada yada) that will follow. But hey, the content of this post is just Bulldozer00’s uncredentialed and non-expert opinion, so don’t believe a word of it.


April 14, 2011 3 comments

Prominent and presidentially decorated software process guru Watts Humphrey passed away last year. Over the years, I’ve read a lot of his work and I’ve always been ambivalent towards his ideas and methods. On the one hand, I think he’s got it right when he says Peter-Drucker-derived things like:

Since managers don’t and can’t know what knowledge workers do, they can’t manage knowledge workers. Thus, knowledge workers must manage themselves.

On the other hand, I’m turned off when he starts promoting his arcane and overly-pedantic TSPPSP methodology. To me, his heavy, right wing measurement and prescriptive planning methods are an accountant’s dream and an undue burden on left leaning software development teams. Also, in at least his final two books, he targets his expert TSP-PSP way at “executives, senior managers, coaches and team leaders” while implying that knowledge workers are “them” in the familiar, dysfunctional “us vs. them” binary mindset (that I suffer from too).

I really want to like Watts and his CMMI, TSP-PSP babies, but I just can’t – at least at the moment. How about you? It would be kool if I received a bunch of answers from a mixture of managers and “knowledge workers“. However, since this blog is read by about 10 people and I have no idea whether they’re knowledge workers or managers or if they even heard of TSP-PSP or Watts Humphrey, I most likely won’t get any. 🙂

Defect Diaries

Once again, I’ve stolen a graphic from Watts Humphrey and James Over’s book, “Leadership, Teamwork, and Trust“:

According to this performance chart, the best software developers (those in the first quartile of the distribution) “injected” on average less than 50 bugs (Watts calls them defects) per 1000 lines of code over the entire development effort and less than 10 per 1000 lines of code during testing. Assuming that the numbers objectively and fairly represent “reality“, the difference in quality between the top 25% and bottom 25% of this specific developer sample is approximately a factor of 250/50 = 5.

What I’d like to know, and Humphrey/Over don’t explicitly say in the book (unless I missed it), is how these numbers were measured/reported and how disruptive it was to the development process and team? I’d also like to know what the results were used for; what decisions were made by management based on the data. Let’s speculate…

I picture developers working away, designing/coding/compiling/testing against some requirements and architecture artifacts that they may or may not have had hand in producing. Upon encountering each compiler, linker and runtime error, each developer logs the nature of the error in his/her personal “demerit diary” and fixes it. During independent testing: the testers log and report each error they find; the development team localizes the bug to a module; and the specific developer who injected the defect fixes it and records it in his/her demerit diary.

What’s wrong with this picture? Could it be that developers wouldn’t be the slightest bit motivated to measure and record their “bug injection rates” in personal demerit diaries – and then top it off by periodically submitting their report cards lovingly to their trustworthy management “superiors“? Don’t laugh, because there’s quite a body of evidence that shows that Mr. Humphrey’s PSP/TSP process, which requires institutionalization of this “defection injection rate recording and reporting” practice, is in operation at several (many?) financially successful institutions. If you’re currently working in one of these PSP/TSP institutions, please share your experience with me. I’m curious to hear personal stories directly from PSP/TSP developers – not just from Mr. Humphrey and the CMU SEI.

The Boundary

March 29, 2011 6 comments

Mr. Watts Humphrey‘s final book, titled “Leadership, Teamwork, and Trust: Building a Competitive Software Capability” was recently released and I’ve been reading it online. Since I’m in the front end of the book, before the TSPPSP crap, I mean “stuff“, is placed into the limelight for sale, I’m enjoying what Watts and co-author James W. Over have written about the 21st century “management of knowledge workers problem“. Knowledge workers manipulate knowledge in the confines of their heads to create new knowledge. Physical laborers manipulate material objects to create new objects. Since, unlike physical work, knowledge work is invisible, Humphrey and Over (rightly) assert that knowledge work can’t be managed by traditional, early 20th century, management methods. In their own words:

Knowledge workers take what is known, and after modifying and extending it, they combine it with other related knowledge to actually create new knowledge. This means they are working at the boundary between what is known and what is unknown. They are extending our total storehouse of knowledge, and in doing so, they are creating economic value. – Watts Humphrey & James W. Over

But Watts and Over seem inconsistent to me (and it’s probably just me). They talk about the boundary ‘tween the known and the unknown, yet they advocate the heavyweight pre-planning of tasks down to the 10 hour level of granularity. When you know in advance that you’ll be spending a large portion of your time exploring and fumbling around in unknown territory, it’s delusional for others who don’t have to do the work themselves to expect you to chunk and pre-plan your tasks in 10 hour increments, no?

Nothing is impossible for the man who doesn’t have to do it himself. – A. H. Weiler

Mangled Model

March 28, 2011 Leave a comment

In their book, “Leadership, Teamwork, and Trust: Building a Competitive Software Capability“, Watts Humphrey and James Over model a typical software project system via the diagram below (note that they have separate Quality Assurance and Test groups and they put the “project office” on top).

Bulldozer00 would have modeled the typical project system like this:

Of course, the immature and childish BD00 model would be “inappropriate” for inclusion into a serious book that assumes impeccable, business-like behavior and maturity emanating from each sub-group. Oh, and the book wouldn’t sell many copies to the deep pocketed audience that it targets. D’oh!

A Free Pass

March 26, 2011 Leave a comment

In a culture of blame, and its Siamese twin, fear, any non-manager group member who consistently asks tough questions and points out shoddy, incomplete, ambiguous work becomes a group target for retribution. This defensive peer group behavior is a natural response to redirect attention away from the stank and to squelch criticism. The funny thing is, managers in CCHs are given a free pass to ask tough questions and criticize without fear of retribution. It helps that managers don’t produce any work products that can be scrutinized by DICsters – if they wanted to. Even if managers did pitch in by leading by example, most DICkies wouldn’t point out flaws because of……. fear of downstream retribution.

Ironically, because of the hierarchical mindset ingrained into all members of a DYSCO, and even though bad managers don’t have to worry about being tarred and feathered by the DICforce, most managers at the workface are incapable of asking the tough questions. Watts Humphrey summarizes this managerial shortcoming nicely:

However, as (Peter) Drucker pointed out, managers can’t manage knowledge work. This means that they cannot plan knowledge work, they cannot monitor and track such work, and they cannot determine and report on job status. – Watts Humphrey & James Over

Cultures of blame and fear of retribution go hand in hand with command and control hierarchies like peas and carrots, Jenny and Forrest. To expect otherwise is to be delusional.

Apples And Oranges

March 25, 2011 2 comments

In “Leadership, Teamwork, and Trust“, Watts Humphrey and James Over build a case against the classic “Test To Remove Defects” mindset (see the left portion of the figure below). They assert that testing alone is not enough to ensure quality – especially as software systems grow larger and commensurately more complex. Their solution to the problem (shown on the right of the figure below) calls for more reviews and inspections, but I’m confused as to when they’re supposed to occur: before, after, or interwoven with design/coding?

If you compare the left and right hand sides of the figure, you may have come to the same conclusion as I have: it seems like an apples to oranges comparison? The left portion seems more “concrete” than the right portion, no? Since they’re not enumerated on the right, are the “concrete” design and code/build tasks shown on the left subsumed within the more generic “Product Development” box on the right?

In order to un-confuse myself, I generated the equivalent of the Humphrey/Over TSP (Team Software Process) quality process diagram on the right using the diagram on the left as the starting reference point. Here’s the resulting apples-to-apples comparison:

If this is what Humphrey and Over meant, then I’m no longer confused. Their TSP approach to software quality is to supplement unit and system testing with reviews and inspections before testing occurs. I wish they said that in the first place. Maybe they did, but I missed it.

%d bloggers like this: