Hal Stern's thoughts on the economy, software, services, technology, and snowmen. Hal Stern: The Morning Snowman

Our historical models for commerce and corporations rely heavily on the notion that distribution and compensation happen at the same time. Buy a software package, you get the license keys (or the installation CD) at the time and/or place of purchase. Buy an audio CD or a DVD, and the goods and money are exchanged at the point of retail distribution. A variety of economic literature exists looking at how and why corporations exist in this model, primarily arguing that corporations lower the transactional cost of commerce. Ronald Coase won a Nobel Prize for this work, and it makes sense -- the companies most efficient at creating products make money; the companies like Wal-Mart most efficient at distributing products also make money.

What happens when distribution and compensation are not tied? What is the economic model for broad, widely accessible distribution with compensation derived from places other than the point of sale? The model isn't so foreign in the consumer space -- this is how pre-paid phone cards work (the wide distribution of phones creates a market) and it's how musicians get compensated for having their music played on broadcast radio. The ASCAP/BMI royalty system charges radio stations a "tax" based on their financials, and the royalties are distributed to artists based on a sampling of what's been played. Distribution is free and wide, compensation happens periodically with statistical models for a true-up of the sing to ka-ching to bling cycle.

Now modulate the accepted commerce models by a low-cost, ubiquitous and high-bandwidth set of networks, one of which is the wireless service I'm consuming sitting in Newark Airport at this early hour. The Internet provides a distribution channel that is significantly lower cost than any real-world retail channel, and with greater reach, greater precision, and greater transparency than the tiered worlds of planes, cranes and country song refrains (about truck drivers). If you can represent your product as bits -- music, movies, images, software, advice, contracts, endorsements -- the notion that you'd tie compensation to distribution is flawed.

The beauty of the broad penetration of Internet distribution is that it fosters the creation of communities of interest. This is what makes eBay tick; it's what makes open source projects work. The primary driver for creating open source software is to create a broad distribution. No different than the wireless companies giving away "free" handsets that are locked to their phone services.

Compensation is independent of distribution in this world. If you want to monetize open source projects, you can charge for their support (viz. Red Hat), create value-add on top of the distribution (a la Oracle's database running on Linux), or create additional services of interest to the community that are related to the distribution (like Sun's compute grid and the Sun Developer Network). Economic models are just as Coase-y in this world -- the companies best at driving down transaction costs for delivering value-add or servicing the needs of the community will thrive.

The bottom line is that open source software isn't a product, it's a mechanism -- it creates distributions. How companies monetize the distribution will depend on how well they address the care and feeding of the community. The transactional sales compensation model for a lot of bits is declining (despite the protests of the major media players, despite the fact that they've benefitted from the "alternative" non-point of sale models like ASCAP and BMI for decades). Welcome to the world of broad distribution.

And today it has a new player: OpenSolaris. Nearly 1,000 pages of content have been provided via blogs.sun.com, kick-starting that care and feeding process.

Hello, world.

In the weeks leading up to the public unveiling of OpenSolaris, I've had many heated conversations about how software development is and will be affected by large-scale open source projects. There are no golden rules or absolute truths here; for every example I provide there are counter examples, which simply means that innovation is hard to predict.

The underlying question for these discussions came from customer queries about OpenSolaris: If we open source our precious operating system, what will we do next? The answer is what it has been for 23 years -- we'll keep inventing new things. The reason is that Sun, like any technology company with a reasonable R&D budget, expects results from the investment in basic research. The "R" component funds new ideas in operating systems, performance, networking, and interoperability. I call this initial innovation. Not just improvements to existing systems, but fundamentally big new things like Dtrace. Because the number of world-class basic computer science researchers is small, and because most of them like to work in funded companies or universities, quite a bit of the initial innovation comes from those same places. This is not peculiar to computer scientists or chip designers; it's true for brain surgeons and drug researchers and other initial inventors.

If you're a brain surgeon, the initial innovation from the major teaching centers is shared in journals and conferences; the rest of the grey matter world can them improve, implement and fine-tune the methods and procedures that come out of the labs. This is the "D" part of R&D, taking an idea and developing it into a product with market stamina. I call this incremental innovation, and it's just as critical and important as the intial innovation, but it happens on a larger scale, in more distributed fashion, and often more informally. Every time you go to see a doctor who keeps up with his or her field, you're (silently) thankful for incremental innovation.

What if you're a software developer? There are ample opportunities for incremental innovation in the operating system, web infrastructure and dynamic language areas. This doesn't imply that initial innovation can't ever come from the community; surely languages like perl prove the point. One or a few developers will always be capable of developing a critical mass of game-changing software, but the bulk of initial innovation is likely to come from funded "R" at companies like Sun and Microsoft and universities with computer sciences programs. With that innovation being opened to developer communities, more of the "D" will be driven by incremental innovation on top of the open source distribution of those initially interesting bits. Even if you're not a rocket scientist, you can at least emulate a brain surgeon.

So what's next for Sun after OpenSolaris? More innovation. More community involvement. More development. How will Sun remain competitive in the operating system space once the source code is opened up? Same way we have - more innovation. How will the users benefit? From more innovation on that source base, whether it's dozens of incremental improvements or a few large-bore clever ideas. Today's event isn't the terminal point, it's the first point in the trajectory of several innovation paths.

Thanks to Bryan Cantrill, Solaris Kernel Engineer, Claire Giordano, aforementioned queen of OpenSolaris things, and Patrick Kerpan, CTO of Borland, for their thoughts, cajoling and insights in this area. Yes, CTOs can be non-confrontational and share ideas, but if we resort to calling each other "doctor" it's usually not with a straight face.

Logic is a funny thing when modulated by the trade press and marketing. We are used to dealing with simple statements in which one predicate is either true or false because there's a singular definition of "true" for the question that has been phrased. Either I'm wearing a red shirt or I'm not wearing a red shirt. Unless you debate the definition of "red" and "shirt", this one is pretty clear. Logicians refer to this as the law of excluded middle and that works fine provided you can be clear about your definitions of truth.

The middle ground excluded by nice, clean logic reappears when there are several possible true states "in between" the two extremes posited in your reasoning. And this is why some of the publicly promoted logic surrounding Sun, open source, OpenSolaris, and Linux has been so flawed. "Sun has open sourced Solaris, so it must hate Linux" is equivalent to "Either Sun does OpenSolaris or it embraces Linux". The flaw here is that the excluded middle in that (highly flawed) preposition contains a variety of other truths, involving Sun endorsing - through support, sales, and source code contribution - a variety of operating system platforms.

Opening the source code for Solaris doesn't mean we dismiss Linux. Using the CDDL license doesn't mean Sun finds the GPL distasteful or wrong. Starting today, there's a world in which OpenSolaris and Linux (and the Apache Foundation and now the Fedora Foundation) represent multiple, independent points of truth. No excluded middle. CDDL, GPL (and MPL and APL and BSD licensing) exist as multiple truths in a twisty maze of legal passages. These are ecologies; they exist in nature, in communities of practice in medicine, and in electronic form.

Ecologies stimulate evoluation -- innovation in our technical world. I'd rather have multiple, vibrant communities of operating systems and the legal eagles governing their distribution than a world with one dominant OS be it Windows, Linux or any other Unix variant. Variety isn't just the spice of life, it's a critical ingredient for its evolution.

Starting today, any average developer can access the OpenSolaris distribution. The middle ground isn't shrinking or excluded; it's just about to be invented.