A couple of weeks ago Steve Yegge, a Google engineer and former Amazon employee, wrote a long, impassioned, internal blog post for Googlers only -- and accidentally made it public.
It's an extraordinary document. Most of the reaction in the blogosphere has focused on Yegge's stinging critique of Google+ (which, ironically, was what he was using when he accidentally opened the rant for public view) and his characterization of Amazon CEO Jeff Bezos as "an infamous micro-manager" who "makes ordinary control freaks look like stoned hippies."
But there's a lot more to Yegge's trenchant analysis.
Yegge's main focus is on something Amazon did very well -- that is, enforce SOA (service-oriented architecture). According to Yegge, in 2002 or thereabouts, Bezos mandated that all Amazon teams had to expose their data and functionality through service interfaces, and that those teams needed to use those service interfaces to communicate. Period. Or else.
It was a forced march to SOA that took several years. But the result, according to Yegge, is that Amazon is now an extensible, programmable platform -- and that Amazon Web Services could not have been built without it.
Having covered SOA intensively for several years, I can tell you that the kind of transformation Bezos was able to pull off is quite rare. SOA is one of those initiatives that's for the greater good: Going forward you can build applications using shared services instead of one-off code -- and when you need to improve what a service delivers, every application that uses it is gets the benefit of that improvement. But in the short term? Typically, SOA doesn't do the service owner much good.
That's why so many organizations struggled with SOA. When each fiefdom in an organization has its own agenda, getting all of them to march to the same SOA tune is not easy, especially when the benefit takes a while to arrive.
When I read Yegge's post, I was reminded of George Glass, chief architect at BT, who also used a hardball approach to roll out SOA in his organization. If developers, instead of using service interfaces to access existing functionality, duplicated that functionality by building it from scratch, they would lose part of their compensation. And guess what? BT ended up with a broadly successful SOA implementation.
Yegge quipped that Bezos was "a regular... well, Steve Jobs, I guess. Except without the fashion or design sense." Job's is another guy who broke a few eggs to make stuff happen. And the recent Jobs biography everyone is talking about revels in recounting the dark side of that style of leadership.
I've spent a good portion of my life wishing and hoping that people would recognize their enlightened self-interest. "If we all chip in on this initiative, the company will benefit tremendously in the long run and we'll all get a piece of that benefit." The response tends to be: "Are you kidding me? I gotta make my numbers this month." Unless someone is holding their feet to the fire.
Today, IT is under tremendous pressure, and the core principle of SOA -- shared services rather than siloed projects -- must now be applied to infrastructure. Or else, in many organizations a persistent inability of IT to deliver successful results on time may turn Nick Carr's "IT doesn't matter" prophesy of years ago into reality. But what will it take to get IT organizations to stop spinning long enough to invest the time and effort on transformational change?
I am not a fan of the strong-arm leadership style. But given the choice between that and IT slipping into irrelevancy, I am forced to choose the former. The fact is that SOA flat out failed in many organizations; at Amazon, it triumphed. Steve Yegge has done us all a service with yet another vivid example of what it often seems to take for IT to effect real, meaningful change in organizations.
This article, "What it takes to change IT," originally appeared at InfoWorld.com. Read more of Eric Knorr's Modernizing IT blog, and for the latest business technology news, follow InfoWorld on Twitter.