September 4, 2004

Microsoft vs. Linux vs. vendor lock-in

Author: Robin 'Roblimo' Miller

One of the big reasons enterprise IT users eye Linux hungrily is that it offers a chance to break the ties that bind them to Microsoft. It's bad business to rely on anything, whether it's a truck or an operating system, that can only be repaired by one supplier. Indeed, one of Microsoft's big sales advantages when PCs were just starting to become popular was that it freed companies from hardware vendor lock-in by offering an operating system that would work with PCs and peripherals made by many companies. Now we need to ask how we can keep from getting locked in by Linux vendors. There's simply no getting around the fact that, from the vendor's point of view, customer lock-in is wonderful no matter what product that vendor sells.I used to run a small limousine service. From the late 1980s through the mid-90s I bought Lincoln Town Cars, both stretched and unstretched, not because they were the most sophisticated vehicles out there but because parts were available for them from many suppliers at low prices, and because almost any mechanic could fix them without special tools. Before Ford's dominance, General Motors was the preferred transport car supplier; Cadillacs were the standard limousine or executive car, with Chevrolet Caprices and Impalas considered the best value for taxi service.

When you put over 100,000 miles a year on a vehicle and keep it for five years or more, your total cost of ownership (TCO) is determined more by maintenance expense than by original price. "Why not Mercedes?" you might ask. Because parts and service were so much higher for Mercedes than for American vehicles that the Mercedes durability factor wasn't enough to overcome the difference. It cost less, per mile, to replace the engine and transmission in a Lincoln every 200,000 miles than to replace those units in a Mercedes every 300,000 miles.

I'm talking about the U.S., of course, where Mercedes dealers have a fair amount of lock-in on parts and service. It was a joke in the industry that alternators and other commonly-replaced parts tripled in value when they crossed the Atlantic. Taxi and limousine operators have associations and conventions just like everyone else. And at those conventions, operators from Europe and the U.S. and other parts of the world get to meet and compare notes -- and repair cost data -- and we found that while Mercedes and other European car parts were high as bleep in the U.S., the same was true of American car parts in Europe.

Whether it's a limousine, truck or computer, smart business people always prefer buying brands that can be maintained at a reasonable cost. European limo operators tend to buy European cars -- and American limo operators buy American ones -- for exactly this reason. And Microsoft has no trouble whatsoever selling corporate computer users the idea that as long as they use Microsoft's one-size-fits-all software, they can free themselves from dependence on one hardware manufacturer.

If you were Novell, wouldn't you want locked-in customers?

Of course you would. Your dream customer would use SUSE on desktops, SUSE and Netware in the server room, and Red Carpet to update everything. Note that neither the word "Linux" nor the term "GNU/Linux" was used in the last sentence. From a corporate marketing perspective, Novell would be just as happy selling WhoozisOS as Linux, and in any marketer's dream world the customer not only buys products from you but comes back regularly for parts and service.

I'm not picking on Novell; you could say the same about Red Hat, Sun, Oracle, Computer Associates, Progeny, IBM, Joe's Computer Repair, or any other company that sells or services computer hardware or software. On a more personal level, we've all heard tales of (or had experience with) programmers who didn't document their code well on purpose -- as a job security measure.

Everybody wants regular customers. There's nothing wrong with that. The question is how you go about getting them. Making it hard for them to go elsewhere once they have your stuff installed is a popular way to do this in the computer industry. There's no reason we shouldn't expect companies that base their business on Linux to do the same.

Look at the enterprise-level commercial software being released for Linux. Often you find binaries available only for Red Hat and SUSE, and if a commercial Linux program is packaged for only one distribution, it's almost always Red Hat. If you're choosing a Linux distribution for a company that's dependent on proprietary software of some sort, you may be forced to choose Red Hat or SUSE even if you personally prefer Debian.

I know, I know. Everyone should use free or open source software. But in real-world businesses, that's not always an option. Many companies have built their IT structures around proprietary billing, ordering, payroll or Enterprise Resource Planning (ERP) packages. A company in this situation can only move to Linux if it's primary software vendor(s) do Linux ports. Those ports are likely to be released only as binaries, and the vendor(s) may offer support for only one or two Linux distributions.

Now put yourself in the (proprietary) software vendor's place: Which Linux distribution would be your number one choice? What would be number two?

Okay, I just opened us up for a flood of "Gentoo, of course!" comments, but in real life a proprietary software vendor is looking at marketing viability and IT management mindshare penetration, not the distribution that has the most ardent fans. And that brings us right back to Red Hat as number one with SUSE number two.

It's going to be interesting to see how distributions besides Red Hat and SUSE fare in the enterprise. Will large company IT departments ignore them? Will Sun manage to get enough desktop deals for its Linux entry that it will be able to make a credible push onto corporate desktops? Will Debian -- and Debian-derived commercial distributions such as Xandros, MEPIS, and Linspire -- be cut out of the enterprise purchasing equation?

Without support from enterprise-level commercial software vendors, it may be hard for any distribution other than Red Hat and SUSE to find a favorable reception from enterprise IT buyers; especially in light of modern IT management's justifiable desire to standardize on as few operating systems as possible. And it's going to be especially hard for volunteer-run operations like Debian to compete with commercial Linux distributors that have substantial marketing budgets when it comes to the corporate purchasing world.

"Free as in beer" doesn't help much when trying to talk a corporate person into choosing a volunteer-run distribution over one with plenty of engineers on staff you can call for help -- for a price. The cost of an operating system is trivial compared to the other costs of running a mission-critical server. I may be wrong, but I suspect that aside from companies that have dedicated IT staffs in-house who show a strong preference for a distribution other than the two major commercial ones -- and can demonstrate their ability to support that distribution without help from outsiders -- we are going to see Red Hat and Novell increase their dominance of corporate server rooms.

Where does this leave Mandrake?

The last time I sat down with Mandrakesoft CEO François Bancilhon, just a few weeks ago, he told me Mandrake was making a major enterprise sales push in Europe and was doing quite well with it. Mandrake has great installation and admin tools -- some people think they're superior to SUSE's YaST -- and a huge user base. Although Mandrake hasn't gotten as much press lately as it did a few years ago, it's still one of the most popular Linux downloads there is.

This popularity means there are plenty of Linux users familiar with Mandrake, many of whom work for companies that might consider migrating some or all of their operations to Linux. This gives Mandrake a huge grassroots marketing force. Add to that the fact that Mandrake is generally attractive on the desktop, and is user-friendly enough that a non-technical manager can figure out how to use it without a great deal of help, and Mandrake may end up coming from behind and surprising its larger competitors.

Besides the mindshare Mandrake has as a result of all those free downloads, it has another advantage over most distributions when it comes to corporate marketability: RPM binaries designed to work with Red Hat or SUSE work with Mandrake either out of the box or with only minor configuration changes.

Most competent Linux sysadmins and consultants can dance between distros like Gene Kelly dancing between lamp posts in Paris. Despite all the "Oy! Linux is doomed to fragment like a dropped porcelain bowl!" wailing that crops up from time to time, the various distributions all have the same basic characteristics, and a high-end enterprise sysadmin will want to install software from source instead of from packaged binaries anyway. So even if Mandrake and the others never make as much headway in the corporate marketplace as Red Hat and Novell, there will still continue to be plenty of choice available for Linux users.

It's natural for big companies to deal with other big companies. That's why HP and IBM are always going to prefer the biggest corporate-oriented Linux vendors, and why big companies tend to buy their computers from big hardware vendors. This is the way of the world. Even if Linux attains total supremacy in corporate server rooms, the "big loves big" equation won't change.

But that leaves a whole lot of not-so-big businesses and a whole lot of independent consultants, small independent systems vendors, and other IT companies that are rapidly learning how to make a living with Linux. And for them, the idea of being able to clone an existing server without going through a licensing routine is often reason enough for a switch to Linux. Microsoft's site license plans don't do small companies much good; they're still subject to software audits any time a disgruntled ex-employee raises suspicions about them -- and even if they're straight with Microsoft another proprietary software vendor might still decide to makes their lives rough.

Small companies need free/open source software

This could be a problem with proprietary software running on Linux, too -- and is a good argument in favor of small companies completely avoiding proprietary software.

Indeed, to get the greatest possible choice of IT suppliers, a small company should use nothing but open source software, nothing but generic hardware, and should insist from day one that every bit of its setup be documented thoroughly. These steps give it the greatest possible flexibility when it comes to selecting outside IT help -- and therefore the greatest bargaining power with its IT suppliers.

In the end, proprietary or open source, any sane computer industry vendor wants to make it hard for customers to leave or get services elsewhere, just as a car dealer loves to be your sole source of parts and service for its products.

At the same time, any sane corporate (or personal) computer purchaser wants to have as many sources as possible for hardware and software maintenance and update services that typically cost more, in the long run, than the original purchases.

Windows has an advantage over Linux here because there are so many people with Windows skills that almost every business, no matter how small, has at least one or two people on staff who can handle basic sysadmin tasks like installing and deleting software. And if that person gets in over his or her head, there are countless computer repair shops and consulting services that can help. Linux expertise is not yet that widespread. There aren't three pages of Linux service vendors and consultants in the local phone book, so there is at least the perception among small business owners that servicing Linux is like getting a Mercedes repaired: It is going to take longer and cost more than maintaining an "anyone can fix it" Ford.

In all the furor over commercializing Linux, and all the efforts by large hardware and software vendors to market Linux products, it's easy to ignore the students and computer hobbyists at the heart of the Linux movement. They are not a group worth "marketing to" since they tend to stick to to free (in both senses of the word) software whenever they can. But these non-customers are absolutely essential to the spread of Linux in the commercial realm, because they are the ones whose presence, everywhere, will make businesses feel comfortable using Linux instead of the ubiquitous Windows.

In other words, instead of "world domination," Linux needs "strip mall presence." We need to see lots of small businesses offering Linux computers, software, service and support. We need to see smiling penguins in store windows and ads in the Sunday paper that say, "Now you can have the same operating system that runs some of the world's most powerful supercomputers on your office desk." We need highly visible competition in selling Linux to retail customers, either in dedicated stores or next to white box Windows computers in generalized computer sales environments.

The entrepreneurs who will open these stores and start the consulting businesses and run the local "Why you need Linux" ads are today's computer hobbyists and Linux-loving students. They are the key to making sure business owners recognize that when they choose Linux, they are not going to be locked into a relationship with a single consultant or employee. And this freedom to choose between many suppliers will do more than anything else to help spread Linux among small business users.

Click Here!