TCO, Firefox, Open Source, and the Release Engineering Blues

Getting Dan Kusnetzky, Program Vice President, System Software, Enterprise Computing Group, at IDC to sit a minute and talk isn't easy. He's usually putting on the frequent flyer miles for corporate executives all over the country. But sometimes you just need to grab a cup of coffee between the Apple confab and exec fests (I keep thinking Mini Mac like MiniMe in Austin Powers, or like a small Big Mac- not Mac Mini like the marketing folk want. Do you think anyone noticed how easy it was to swap the name before they decided on it?)

We continued a discussion begun with a Boston Globe quote of Dan's on firefox and open source software - "'A number of things are causing people to give open source a look. One is the security problems that are almost a daily concern in Windows. . . . There's also the cachet. Some people think it's cool to tell their friends, 'I'm using Linux." On this both Dan and I agree. He's right homing in on security as an issue (I'm using firefox myself), and I guess it's cool, even though we first launched an open source BSD OS years ago (remember - 386BSD was the first), so firefox is just a good tool to me - not a religion.

But the article goes on to state "But there are impediments to ordinary PC users embracing Linux. Kusnetzky said thousands of software applications available on Windows don't run on Linux, including popular personal finance programs like Microsoft's Money or Intuit Inc.'s Quicken." Well, I'm not sure if apps are as critical as Dan thinks they are, but I must admit that inertia is on his side here. Still, given the rapid acceptance of firefox, I think we don't want to set our expectations too low.

Most of the people I know use open source products in addition to Windows products, such as openoffice and firefox. In fact, most apps in an office are now available in some form (some good, some not so good) on other platforms. Or you can run apps in a windows emulator nowadays - it is possible, but not quite there yet.

But I don't think the problem is with open source per se. I think the real problem is that no one takes release engineering seriously - and this bane afflicts commercial and open source software. Even our most mediocre 386BSD releases underwent more testing and had far fewer issues than many of the releases (both commercial and ad hoc) that I work with today. The patch mentality is used in lieu of good sales and support - if they whine, give them a hack. It won't cost you much and it will make them happy. This isn't right.

The problem with the patch and update assumption is that it just isn't cost-effective in the long-run. The amount of effort in maintenance and coordination of these little bits of code begin to override the real product roadmap goals and obligations. Sales spends time in talking patches, not products. People get used to something (even of lousy quality) for nothing instead of paying for something they've been sold. Sales gets cynical because nothing is good enough to sell. Engineering / product management wallows in the legacies of the past instead of designing the future. Customers feel frustrated and locked-in (gotta lock them in with broken software) and take their frustrations out on the company, which hands them a patch, and the cycle continues. And business suffers because they can never get past the mediocrity of stale products that never deliver.

The lost business revenue is real and measurable. But this hasn't been taken seriously yet. Recall that TCO came into vogue to convince people that the cost wasn't just the hardware or software alone - it was the maintanance and downtime that really cost. That's what made enterprise computing take off.

The problem is that most people haven't applied these same rules to open source and commercial companies which now emulate open source. Some like in the German government are pushing open source to deal with the total costs of the windows franchise (openoffice / staroffice solutions) that they claim lower the cost of maintaining an entire city's infrastructure. They claim real cost recovery here. And in Germany, where brand value has worked backwards in many cases in recent years, moving away from Microsoft products, partially because of brand but more so because of total costs, even putting up with real serious (frustrating) issues to migrate just to recover these costs - well, Microsoft can't fight it. Asian governments are watching this closely.

TCO has to take into account some issues that enterprise TCO calcs doesn't - like release engineering / consistency / patch cost of operation. Enterprise systems were so customized that these issues were a part of the total installation. But of course, it was these very costs that caused Stanford, for example, to outsource to India all their telcom and accounting systems maintanance last year.

I think we've just mined out this "let the customer do the work" patching and hacking tricks to lower costs. The race to the bottom has basically hit bottom, and is actually increasing costs. It's time some real rigor was applied to quantify this objectively.