Actually, you are asking two separate questions, both of which are equally valid.

The first question is: How does one build an industry?

The answer, as Lee Iverson stated so succinctly, is "simple standards and free software". The HTML standard, coupled with the Mosaic browser, did indeed create a new industry, defining a new defacto standard upon which communications have become increasingly based. Yahoo did it, as well.

The second question: How does one create a self-sustaining *company*? (In particular, one that operates within that industry.) As you say, this is fundamentally a question of business models. What business model makes sense? What it is the value proposition that produces revenue?

I have to confess to being almost totally mystified on this point. I *still* do not understand the Netscape/Yahoo/RedHat business models. It's not clear to me how they go about making any money at all, much less enough to support a large organization. There is manifestly *some* model that makes sense, though.

I started investigating question at the after-colloquium party. Since VC folk have a great nose for sniffing out business propositions, I decided to ask "what open source business models are fundable?" (This may not be the best way to express the question, since both Yahoo and Netscape started with college kids giving away their software & services. But I'm thinking that somebody had to invest something, at some point, before they were able to form a company. Eugene may have better information, here.)

These are the models I've seen so far [2] To this, I add one theoretical possibility [Horse Race model] that I have never seen occur in practice, but which might be feasible:

1. The Education and Services model

   The software is the key product. You give it away, and make your money providing education and services (such as prioritized development). That's the Red Hat model. I'm not sure it's a huge winner, but so far it seems to be at least reasonably effective.

2. The Derivative Service model

   The software is something you use to provide a valuable service. You have some other value-add with respect to that services that makes it a viable business proposition. VC loves open source in that model, because you get a whole world of developers helping you to refine the software you use to make money. For example, you might provide payroll services to small companies. The payroll software would be free, and large companies might decide to use it. But most small companies will still pay to outsource the operation, for lack of inhouse manpower to take over the job, so your fundamental business proposition is unaffected.

3. Horse Race model

   In this model, a bunch of VC types get together to have a horse race. They each back their own "pony" (individual company). But they pool their resources to create the "race track" (base of open source software they can build on).

Companies agree to standards (with a minimal investment -- time on the part of some employess and possibly a membership fee to the organization to cover adminstrative cost). The standards that result produce the race track, and the company products are the ponies.

In the client/server domain, life gets really interesting. The best situation to be in, by far, is to be producting the server (the platform) and let lots of other companies compete to build the clients (applications). That was the ground Microsoft took so successfully. Then they migrated into applications, as well, which burned a host of application providers.

If you can get to that point, that is ideal. Another alternative is to give away a server (the racetrack) and a minimal client (a seed pony), and let client producers compete with better and faster ponies.

On the other hand, client software is notoriously a low-margin business. (The ideal price is free.) So maybe the race is around producing better servers. You provide a minimal server for a seed pony, as well as a minimal client. The race is then to provide the best possible server, where better/faster clients help to increase the winner's prize (i.e. the market size).

In this model, a open source versions of both server and client software "builds the racetrack". The risk is that no one comes. But if they do, the "gate" (winners prize) becomes the market estimate -- what customers will pay for software that runs on that track. People then place their bets on different ponies (buy stock). That rewards the folks who have ponies.

The folks who built the racetrack, therefore, are only rewarded if they also have a pony in the race. When pioneering a new industry, the risks are greater -- you need to develop freely available versions and hope they catch on at the same time that you need to develop a proprietary version. So your investment costs are higher. Plus the risk increases because if *either* project fails, you lose. Finally, unlike investment in a new potato peeler (which you know people use) investing in a new industry runs the real risk that it never gets accepted. (If no one goes to the racetrack, having the best pony in the race produces no reward.)

The risks are real. If there were some way to create a consortium of VC folk to build the open source pilot, the risks could be minimized for any one participant. But there are still two major reasons for investing in such a project, even if it must be a solo venture:

1. The rewards are huge.

   IBM made out like a bandit because it created the computer industry, and then proceeded to dominate the hell out of it. When you succeed in such a venture, the rewards are massive.

2. We can't afford not to.

   If our planet and our very survival as a species depends on solving the critical, complex problems that confront us, then we cannot afford to ignore technology that promises to augment our collective intelligence. Win or lose, it is an attempt we must make.

   
   

Eric Armstrong
eric.armstrong@eng.sun.com