My six-year-old son loves to experiment. It usually involves a watery mess, usually in the kitchen or bathroom, and the results are often mixed, though he seems to get a lot out of it.
Sun Microsystems is into the second year now of an experiment around software economics, partner ecology development, and enterprise account penetration. Sun's bold and imaginative approaches -- open-source more and more software, make much of it "free," and encourage a subscription pricing model for the non-free aspects -- also appears messy, but I'm still not sure Sun is getting much from it.
The Steward of Java is now making all software under the Sun free and open source, except Java. I'm pretty sure I understand why Sun is holding on tightly to Enterprise Java as those world-changing set of specifications settle It's hard to hold on to a long-term hardware differentiation for most mainstream computing any more. into a trajectory toward maturity and legacy, just another layer of middleware to manage. I guess I also understand why Sun is attempting to re-make open source code distribution in its own image. And I also get that many enterprise users and buyers want a unified and moderately integrated software stack, from soup to nuts, with one throat to choke.
It's just that, like when my son mixes water, salt, sugar, and LEGOs in a bowl and it still doesn't make pancake batter -- when Sun mixes full stack, open source, and subscription pricing, it still doesn't whip up what the market seems to want to digest.
The key question now is: Does the market really want an open source stack from Sun? Or from anyone, for that matter? Is the Sun half-man (open source), half-horse (integrated stack) software experiment a Centaur for the City Zoo -- a nice oddity to view for free on Sunday afternoons, or a market-dominating hybrid solution that charges into data centers as a must-have alternative to all the other code?
The enterprises and service delivery platform architects at carriers and hosts seem to fall into only a few basic camps when it comes to infrastructure choices. One large camp finds itself with a full, integrated stack under each and every major application. They have picked the application and best performing runtime for those apps first, and then made a variety of decisions for a variety of reasons on platform over the years. They have lots of different stuff, and are in the ongoing process of integration, platform triage and hardware consolidation. Web services helps them a lot, but it's still a barn yard and will probably stay that way for a long time.
There are other camps where they make strategic platform choices and then try and line up as many applications on that stack as possible. These are shops where the operations side has the most power, and developers must line up behind the centralized decisions based on homogeneous stack efficiencies. They tend to like packaged applications a lot. They too are involved with triage and consolidation. They have been, and still are, trending toward Linux and Windows for their "standard" platform, and they know full well that the applications they want will be well supported on these platforms. Nice deal on the hardware, too! The first camp is trending toward becoming the second camp.
And then there is the messier third camp where even if they favor the unification of platform or seek out applications first and stack as an afterthought, they tend to be heavy into customization, leveraging lots of open source components in addition to Linux as an OS, and are more interested in custom development advantages than to the operations side of the house. Here the developers hold sway, the applications are king, and the ops folks must support the applications and developers come hell or high water. Ongoing cost is not so critical because the applications are the business, not just in support of the business.
So how does Sun's approach appeal to these camps? Sun would say it appeals to each on their own terms. Sun says that to woo the developers is to pull through the free and open source software infrastructure, and then probably the Sun hardware, maintenance, and support for a long-term, albeit low-margin, relationship. Alternately, Sun would say to appeal to the ops people on unification and low cost will pull the developers along in those ops-strong shops. And Sun would say that by having open source middleware that runs on Windows and Linux that they appeal to the heavy-customization shops based on integratability and adhesion to standards, as well as direct access to the underlying code. And Sun can now show it has a great deal on hardware, too; lots of bang for the buck.
Sun's stock, however, has been trading at below $4 a share for a long time based on the assumption that these Sun visions are experiments and not sure bets. Wall Street does not strongly value Sun's installed base because they fear that it is still fleeing, and that new accounts that Sun may acquire will be low-margin affairs with little lock-in, with perhaps more fleeing when Intel and Dell catch up, as they tend to do.
So perhaps Sun's grand experiment, despite its various sizes fits all logic, actually doesn't fit any all that well, given the alternatives and rapid rate of commoditization. There is little reason among the basic camps as outlined here that Linux on x86 is not plenty good enough. And that fits in fine with the ops folks, the developers, and the ISVs. Windows is not going away, either.
In fairly short order, hardware becomes a wash; it's hard to hold on to a long-term hardware differentiation for most mainstream computing any more. I'm pretty sure the non-Sun hardware vendors will build on multi-core technology quickly and narrow the Sun advantage on power and footprint. Sun's advantage in power consumption and footprint are short- to medium-term differentiators amid a long-term strategic decision process.
If Sun's logic on appealing to the market with its software experiments has legs, we should see them soon. Sun's cat is now fully out of the bag; no more intrigue and suspense. What you now see is what you get with Sun, totally.
I do hope that the legs we see from Sun's adoption are not the same legs running to high-value customization differentiation for applications and integration that is almost entirely abstracted from the middleware, with little thought to the infrastructure levels at or below Java.
Because if all the essential technology differentiation choices are now at, and will for the foreseeable future remain, with the services, interoperability, and process management levels above the platforms, then even free and integrated won't matter much underneath the applications for most enterprises and hosts. Free, almost free, and cheap are no long differentiators nor wedges into the market for middleware. That card has already been played.