% fortune -ae paul murphy

Rx for Linux: Part 1 - Measurement

Here's my premise: from about 1997 to late 2002 Linux growth looked unstoppable. The mainstream press rejoiced over the myth of Linus, IBM bet the software business on taking over Linux, and serious people predicted that it would gain significant market share over Microsoft in both the server and the desktop markets.

Since then Linux has continued to grow - but the oomph has gone and gains are being made the old fashioned way: by hard work. Linspire's recent announcements are encouraging but things like the potential 300,000 unit install in Indiana are just that, potential and, meanwhile, the much hyped Munich transition is in deep trouble.

Permit me the premise and we get an interesting question: what happened? and, more importantly, assuming we wanted to bring public excitement back to Linux, how would we go about doing it?

I believe that many factors contributed to the loss of momentum but the single biggest one was the withdrawal of mainstream media push as IBM's worldwide support for Linux combined with the SCO lawsuit and Red Hat's negative marketing to reduce their enthusiasm for the myth they'd earlier sold themselves.

I do not believe, in contrast, that the constraints imposed on IT spending right after 2000 had anything to do with this. On the contrary, the open source world of which Linux forms a highly visible part handily beats Microsoft on cost and that, combined with the usual Microsoft behaviours on licensing, anti-trust, and product delays should have boosted Linux installation rates beyond all recognition.

But it didn't happen and I think, on balance, that the most likely reason is that mainstream puffery has been largely removed from the pre-decision process - meaning that people outside IT have stopped pushing us to review the Linux alternative in "platform" decisions and that Linux growth rates have therefore returned to what they would have been had the mainstream press never discovered it.

That would, of course, be difficult to prove even if we had good installation rate numbers -but we don't and that's one of the most basic things I believe have to be fixed if Linux is to form the basis of a competitive small systems software market in the major English speaking countries.

Fixing the counting problem should be trivial: all it would take would be for a major Linux organization or other open source community group to set up a counter and ask all affected developers to install a self reporting option that asks the user if it can send the counter some basic installation information. Ideally the counter would be web queriable and give us, for a few weeks work in getting it set up and some minor operational overheads, the first publically available numbers on open source adoption and use.

A counting service could, of course, also look at doing other interesting things - like serving as a demand registry for products that don't yet exist, maintaining failure-free run-time numbers for open source applications on Linux sites, or collecting information on staffing and deployment strategies among commercial Linux users.

There are people who say we shouldn't try to count installs - in their opinion either things are fine now with Linux growing just as it should be, or they agree that growth has slowed but like that for reasons of their own. Either way, there's obviously lots of room in the house of Linux for their views, but I disagree: without real numbers we're all groping in the dark and I don't see that as a good thing for anyone except Microsoft.

The reason the community's continuing failure to count our successes serves Microsoft well is that Linux can reasonably be seen as a kind of entry level Unix that competes mainly with Microsoft's Windows brand products, not the BSDs or Solaris. Look at Linux deployments and, to the extent that anyone knows, they seem to focus on the same markets, and use the same hardware to run fundamentally the same applications, that Windows does.

In other words, the marketing strategy for Linux commercialisation should be modelled on Microsoft's, not that of the BSDs or Solaris.

Red Hat to the contrary, Linux only competes with Solaris for the jobs that Windows could do too - not as reliably, usually not as cheaply, and arguably not even as well - but functionally the same core jobs at the same levels of complexity and throughput.

So what's the the most important lesson we can learn from Microsoft? that nothing sells like success. Start counting installs and making those numbers widely available, and pretty soon what's recently become largely a stealth phenomenon could start to snowball again.

Tomorrow: part two - on getting licensing and patent issues out of the way.

Paul Murphy wrote and published The Unix Guide to Defenestration. Murphy is a 25-year veteran of the I.T. consulting industry, specialising in Unix and Unix-related management issues.