% fortune -ae paul murphy

Sun's worst enemy: Sun

Last week my accidental avatar, "You already have me registered" received a Sun "Inner circle" newsletter featuring an interview with Richard Green, Sun's executive vice president for software, subtitled When "good enough" is no longer good enough and nominally about "Upgrading from Linux to Solaris."

Please be aware, in reading my comments below, that I have a previous relationship with Mr. Green. Specifically, I talked to him several years ago about having Sun support a full out, 500+ user, Sun Ray demonstration project in which we would do everything in public: with user requests and comments, system logs, budgets -everything- available live for review by anyone with a web browser. The discussion fizzled, and while I don't think this affected my reaction to the newsletter you need to be aware that I might not be entirely honest with myself about that.

In the nature of things staged interviews of the kind reported in this newsletter are heavily influenced by American legal considerations and, of course, the need not to offend top IT managers guilty of stupid decisions in the past - you just don't make a lot of Sun sales telling guys who might be considering Sun now that their previous decisions to buy from IBM were, ahem, cough, "uninformed".

The trick to getting them to come to Sun is to cater to their prejudices while ensuring that the new information you feed them sells Sun but doesn't directly challenge things they're sure of - particularly if those are all wrong.

What makes this especially difficult for Sun is that many of its own people are relatively new to the company and believe much the same nonsense their clients do - meaning that someone in an executive position has to be especially careful to get the balance between telling the truth and sucking up to the customer exactly right, because the company's own troops are going to be weighing every word a lot more carefully than the customers do.

Generally speaking the best way to go beyond merely catering to customer beliefs to both sell Sun and actually be right too, is to pretend that history has no influence and all decisions are de novo: i.e. pretend that customer decisions are exclusively driven from rational analysis of the current and expected future state of things at the time the decision is made - or, in practice, by knowing when to keep your mouth shut.

Unfortunately in this instance Mr. Green doesn't seem to - and, instead, ratifies so many erroneous, but common, customer beliefs that I suspect the overall impact of the document will be more positive for Dell, Intel, and IBM than for Sun.

Here's the opening salvo:

IC: Let's start out by asking a basic question: What was behind the initial appeal of Linux to support the Web tier of applications and services?

GREEN: People picked Linux in the past for a number of very good reasons. First, starting in the mid-'90s, the adoption of Linux was largely fuelled by fast processors. Intel, for example, was producing processors for the desktop at a price-performance level so attractive that it made a lot of sense to start using commodity-based or volume-based application infrastructures. With the rapid growth of the Web tier, a free UNIX-like operating system coupled with inexpensive servers proved pretty attractive.

Second, the same period witnessed the increasing popularity of the open source development model. Having access to source code pushes innovation, and the kernel Linus Torvalds built became a nucleus for more people to add small innovations to Linux similar to the way that snowflakes form a snowball. Finally, after the dot-com bubble, economic realities made good enough technologies, rather than best-of-breed but expensive technologies, quite popular. So for some organizations, a low-cost operating system that performed adequately offered an attractive price-to-performance coefficient.

There's nothing wrong with starting something like this by assuring the customer that his past decision to get into Linux was smart -it almost certainly was- but going on from there to endorse the view that Linux adoption "was largely fuelled by fast processors" produced "for the desktop at a price-performance level so attractive that it made a lot of sense to start using commodity-based or volume-based application infrastructure", isn't.

In fact it conflates multiple issues, applies essentially meaningless buzzies ("volume-based application infrastructure") as a kind of customer unguent, and could be more or less translated into English as "people bought Lintel because it was cheaper, faster, and less proprietary than SPARC/Solaris." Worse, his follow up patronises Linux as some kind of second rate Unix clone while implicitly putting down Sun: "economic realities made good enough technologies, rather than best-of-breed but expensive technologies, quite popular."

All of this under sells Sun in part because this is a senior Sun guy talking, but mainly because none of it's true. On the product side Sun's 450/250 series of UltraSPARC II servers were designed to compete directly with Compaq's Proliant line - and not only did so quite effectively on cost and performance at the time, but ultimately proved to have been far better investments in two ways: first because many are still in daily use and, secondly, because anything that ran on SPARC under Solaris 2.7 or 8 runs equally well on Solaris 10 today.

What he should have said with respect to x86, of course, was that people had lots of small x86 servers sitting idle because Windows licensing and product obsolescence made decommissioning this gear attractive - and adding a free operating system extended that gear's useful life, first for compsci students and others for whom the money meant something, and then, as the software developed and these people moved into businesses, for organizations.

More importantly, the purpose of the "interview" was to get people thinking about moving from Linux to Solaris -and so the question he should have asked himself first was why people choose Linux over BSD - at the time a more established Unix for x86 with more software and better support that also ran faster and more reliably on x86 than Linux did.

The rest of the interview -which I'd urge you to read - doesn't reveal any deeper thought; and what I see as his willingness to run down Linux doesn't reflect well on Sun either - in fact, that attitude probably confirms some beliefs people cite for not buying Sun's x86 products: specifically that Sun isn't serious about either Linux or x86.

Consider, for example, this exchange:

IC: Have perceptions changed about using Linux in the enterprise?

GREEN: I hear a lot of talk emanating all the way from the datacenter floor to the executive boardroom concerning the unexpected challenges and hidden costs of using Linux. In the early days of corporate Linux adoption, many companies felt compelled to hire operating system teams to support Linux, and these groups would take something that was free, and spend substantial amounts of money and time modifying it to meet the specific requirements of the particular enterprise.

While the ability to modify is appealing to many developers - as well as one of the attractive aspects of open-source software - an operating system is an extraordinarily complex piece of technology. And the need to constantly tinker with the operating system necessarily drives up costs and negatively impacts overall productivity.

IC: But companies like Red Hat and Novell offer consulting services for optimising Linux to mitigate some of that risk, don't they?

GREEN: They do, but they don't provide those services for free. Of course, there's nothing wrong with charging a premium for a better product, but many companies have discovered that an OS that had first been presented as being "free" in cost actually incurs costs similar to or exceeding what, say, Microsoft charges for its operating system. Many companies are learning that while Linux is relatively inexpensive to acquire, it comes with high deployment and maintenance costs.

Of course he's catering to his market and most of them are so clueless about how to run Linux that they do buy support contracts for each PC they use it on -while digging themselves even deeper holes by paying people to tinker when they shouldn't - but he's doing nothing to help here. Instead, he's setting up their next round of failures by validating their beliefs about how to manage Unix - failures they'll just as surely blame on SPARC/Solaris in 2010/11 as they're ducking responsibility by blaming Linux now, and exactly as they did after their last attempt to bend Unix to their views failed.

That's the bottom line on what's really objectionable about the message in that newsletter: it's all about the customer massage: patronizing, convenient, shallow, and wrong - and therefore neither good for the customer nor good for Sun.


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