November 9th, 2005
Was OpenSolaris a Mistake?
In my Monday blog a week ago I suggested that Sun’s management could reasonably do an employee buyout by leaving Sun’s consulting services in the company and taking everything else private. That way the consultants could go and become an EDS or Accenture, the shareholders would be better off, and the real network computing company could stop trying to play both sides of the fence in its relationships with its partners and customers.
The very first talkback comment on this came from Anton Philidor. Here’s part of what he said:
Solaris was a major asset. Now they’ve open sourced it. I thought the reason for this questionable move was selling services.
I’m probably not following your argument completely. Wouldn’t shedding services make open sourcing Solaris a (more) obvious blunder?
…
Again thinking simplistically, Sun will not be able to underprice mass market hardware makers. And with Linux and Windows both growing rapidly, the cheap software is not going to be Sun’s either.
So Sun is a company with strengths that apparently no one wants which has an unfortunate tendency to give away the assets that might persuade people that the company can go against the tide.
There are several issues here and I want to focus on this idea that open sourcing Solaris might have been a strategic mistake. Before getting to that, however, I’d like to answer the other points:
- Right now, Sun offers the fastest, and cheapest, x86 boxes around. Prior to the recent introduction of the "Galaxy" (aka "becky boxes") line these were simply the same third party products companies like Dell and IBM ship -and the Sun label on the front didn’t make them any faster, although Sun’s willingness to buy market share did make them marginally cheaper.
The new machines, however, incorporate some advanced design elements aimed at reducing manufacturing cost, increasing reliability, and decreasing power use. As a result they cost less, and work better, than the third party products they replaced - products other companies are still shipping.
- Right now Solaris costs less than Linux - amazing given that Linux is free, but true since Solaris really is free while the people who buy Linux for businesses generally like to pay through the nose for a license in the guise of a support agreement. Further Sun’s x86 boxes run the same cheap Linux software as everybody else - but, and more importantly, the "Janus" run-time means that the SPARC/Solaris machines do too.
How these things will play out as Linux moves to Cell I don’t know, but for right now I don’t see any boats being missed here.
And that brings us to the main issue: was open sourcing Solaris a mistake?
Intuitively my feeling is that open is better, so I’ve never really questioned this. Be aware, too, that Solaris (and SunOS) have always been open in the sense that people with the right credentials and a willingness to sign NDAs have had relatively easy access to both source and key developers. Recognize, therefore, that formally opening the source is really more of a broadening of access than a new departure.
At the practical level, open Solaris is key to the CDDL license and the patent umbrella initiative. That initiative, in turn, is the key tactic in a strategic battle for developer loyalty: remember hardware sales lag developer commitments by a few years, so Sun sales in 2009 and 2010 are really dependent on what they get developers to commit to this year and next.
One of DEC’s (and later HP’s) big mistakes was to cut back on developer support: Sun is doing the opposite, with open Solaris a key component of one specific tactic for that. I can’t see this as being anything but good for the company, but we won’t know for sure until we see whether present developer take-up (which is setting records) really does translate into exciting new applications and therefore hardware and services demand in the future.
That’s the practical argument, but there’s a much more compelling theoretical one too.
Let’s assume that a hardware company’s goal is to sell more hardware. In that situation, keeping the company’s operating systems and major applications code proprietary locks in customers already using it and therefore guarantees at least some replacement sales by keeping the rate of customer defection down. Conversely, opening the code gives customers more choice and thus produces market growth during times in which the company’s products are competitive.
As a result, a proprietary response on code is called for when the hardware won’t be competitive for the next few years and openness is called for when the company expects its hardware to lead in cost performance for the next few years.
Right now Sun’s x86 hardware beats everything else out there - and they see a clear roadmap to rack servers offering four AMD CPUs with four cores each. More importantly, the new USIV+ "traditional" Ultrasparcs are beating Power5 on both price and performance while the company is about to introduce a whole new world of high performance, low cost, CMT/SMP computing.
IBM’s Linux on Cell will be faster and cheaper than CMT on floating point, but the software is much more difficult and they’re late in the field while Microsoft’s nexen is years from making it to the data center - in other words, Sun has clear competitive advantages across the board right now.
In this situation openness on code is clearly called for -and going after developer loyalty the number one mission critical strategic activity.
So, bottom line: is openSolaris a mistake? Nope, it’s business school 101 strategy in practice.
Paul Murphy (a pseudonym) is an IT consultant specializing in Unix and related technologies. See his full profile and disclosure of his industry affiliations.
Subscribe to Managing L'unix via Email alerts or RSS.
| 11/10/05
