Measuring open source adoption the hard way

Microsoft is sponsoring an open source measurement project—one that has inventoried a total of 1300 computers.

But despite the sheer number of projects that want to measure open source adoption, the numbers are bogus, because it's not in the interest of the people who have the data to release it.

The more interesting the customer, and the more that customer deploys, the less they'll talk about it. Banks are especially wary of talking about their software habits. Wall Street becoming Linux stronghold" is still news now, almost ten years after Linux broke investment banks's "no eating in the data center" rule by eating Sun's lunch there.

Now, of course, with a steady drumbeat of legal threats (usually phrased something like, "we just want you to compensate us for our Intellectual Property and work together") Linux and free software users have even more reason to keep quiet. And it's not all patent trolls, who tend to be relatively sane and revenue-maximizing. You get wackos. After all, it's still hard to even pin SCO's CEO down on what he claims customers are infringing. Give one "we use Linux" quote to one news site, and it's "On information and belief" in a lawsuit against you. The publicity isn't worth it.

Jay Lyman at the 451 Group says Microsoft has "changed" but it's a big company, companies change back, and companies have partner companies. Sure, they're not suing their customers, but they bankrolled SCO. The threat is there no matter how much happy friendly working together PR you put on it.

I like hearing about real-world business IT success stories as much as the next media weasel, but if I were on the other side, the policy on chatty IT staff would be Shut the Fool Up. Employees, feel free to ask questions on web sites or mailing lists, but use your personal email address, and any discussion that mentions both a software choice and the company name is off limits.

So, how do you measure how much open source is out there? The obvious choice is a "popularity contest" application that checks the filesystem or the package manager.

Look at the Debian Popularity Contest. Of course, it can only measure what's installed, not what you actually use, unless you decide to mount your filesystems without the "noatime" option, but don't do that. Several companies have released open search search tools as sales tools (like this one at OpenLogic) "Look at all the open source stuff you're running and didn't know it—you need our support plan!"

But people don't run popularity contest applications on production machines. So, as much commentary as the OMG WTF M$ WANTS YOUR INSTALLED SOFTWARE LIST thread has gotten, it looks like a waste of the company's money.

Strangely enough, what might be the best open source measuring tool to come along in a while recently got released by accident. Netcraft's report on bad SSL certificates on bank and other secure web sites. (via LWN). That reveals not the OS on the outward-facing server, but the OS on the internal system on which someone ran a broken version of critical software.

Naturally, it would be irresponsible to publish today which banks are running with a bad certificate. But Netcraft now has data on companies running not just open source, but the squeaky-clean Debian distribution whose software guidelines were the basis of the "Open Source" brand in the first place.

(This might be a good one for Revealing Errors.)

Syndicated 2008-06-19 23:59:38 from dmarti's blog