Lessons from the Layoffs at Linuxcare

by Tim O'Reilly

Linuxcare CTO Responds

Dave Sifry reacts to Tim's article.

The venture capital community and tech press is abuzz about recent layoffs at Linuxcare, the Kleiner-Perkins backed Linux support company.

Boosters of open source argue that Linuxcare's stumble is simply a result of management missteps, trying to grow too fast, and the overall cooling of the market to speculative high tech offerings and to Linux in particular. Observers critical of open source might argue that this event shows the weak commercial underpinnings of Linux, and the difficulty of making money when the software is free.

I draw an entirely different lesson: that the "service" opportunity for open source software requires thinking in a much bigger box. (You don't actually have to go outside the box. You just have to give yourself some elbow room.) Linuxcare's initial business model involved a great deal of reliance on phone-based tech support and other low level services; they are now repositioning themselves for higher-level professional services such as creating private label versions of Linux. They are absolutely right to think bigger. The service opportunity is immense, but it isn't necessarily in the obvious places.

Let me illustrate with a couple of stories.

My all-time favorite open-source derived service business is UUnet, now a multi-billion-dollar division of MCI Worldcom. Rick Adams, the founder of UUnet, has never been particularly identified with the free software or open source movement, and to this day doesn't particularly like to acknowledge its role in his success. Nonetheless, he was the author of several important open source packages and built a huge business based on them. He wrote B News, at one time the most popular Usenet news transport software, as well as the first implementation of SLIP (Serial Line IP), the predecessor to PPP in making dialup internet access possible.

For those of you who weren't around at the time, a brief history lesson: You now think of Usenet as something you get from your ISP. In the early days, it was a completely voluntary operation. You got your "Usenet feed" from a neighbor who was already hooked up. You arranged to dial up on a prearranged schedule to pick up and forward mail and news, mostly using the UUCP (Unix to Unix CoPy) communication protocol. It would often take two or three days for a message to propagate itself across the net. Every news site had to download every news message in any newsgroup to which they subscribed, so this could be a very time-consuming operation, requiring hours of connected dialup, often during off-hours. Users would then access the news from spool files on their local machine. (Under today's modern news servers, connected over the Internet, users can access messages in real time from a remote host, first downloading only the headers of messages in a newsgroup, then retrieving any individual messages they want to read.)

Over time, a number of sites emerged as Usenet/UUCPnet hubs, feeding hundreds or even thousands of other sites. Each of the hubs was connected to the others, forming what was called the Usenet "backbone." Rick Adams was the manager of a site called seismo at the U.S. Center for Seismic Studies, which eventually became one of the world's largest Usenet hubs, and was particularly important because it was the only one in the early years with a connection to Europe. The closer you were to seismo or one of the other hubs, the faster your mail and news got through, since it had to travel over fewer hops to get to its destination.

But there was trouble in this cooperative paradise. As the traffic load got bigger, the phone bills of the hub sites also got bigger, to the point that management couldn't look the other way any longer! Why is our dialup phone bill $100,000 a month? How does this support our mission?

Rick was one of the first to realize that his software needed commercial support. If people were willing to pay for UUCP and Usenet access, he could get the bureaucrats off his back. The commercial ISP was born. (At the time, it might have been called a UUSP, for "UUCP Service Provider"; TCP/IP didn't become the biggest part of the business till years later. UUCP lives on as the UU in UUnet's name.)

Rick first started UUnet as a non-profit corporation under the aegis of Usenix. Over time, he realized that the opportunity was bigger than he thought, and converted UUnet to a for-profit corporation. The Internet took off, and UUnet was a big part of making that happen.

(An aside: O'Reilly's own small part in that history ran parallel with UUnet's. In 1986, I wrote a book, Managing UUCP and Usenet, that evangelized the Usenet and told people how to get hooked up. This was O'Reilly's "underground bestseller" of the mid-80's, the book that more than any other defined our earliest publishing efforts. We became a UUnet customer, and Rick distributed the book to all his other customers. When he started offering TCP/IP services as well, we switched. We realized none of this stuff was well documented either, and so commissioned our 1992 book, The Whole Internet User's Guide and Catalog, which was so instrumental in bringing awareness of the Internet to the masses.)

To my mind, Rick and other early ISPs really understood what service means in a world where the software is free. Rick didn't try to put Usenet in a box and sell it. He didn't try to offer tech support services (though they were certainly part of his offering.) He gave the net "dial tone." He took all of the uncertainty out of access. He realized that in order for Usenet to grow to the next level, it needed a more reliable infrastructure, with a backbone that had the resources to offer a level of service that was good enough that people would be glad to pay for it.

To some people, this might make Rick and others like him enemies of free software, at least those who confuse gratis with libre. After all, UUnet took a service that had previously been provided on a volunteer basis, one of the great outpourings of cooperative goodwill in computer history, and replaced it with something you have to pay for. The old free, cooperative dialup net is dead.

But in fact, while the software was (and still is) free, the service it provided was never free. In the old days, we each paid our own phone bills, and the backbone sites with a lot of traffic paid more than their fair share. Rick basically dipped into the revenue stream that we were already paying to the phone company, and made it his own. Heck, now UUnet is a phone company, or at least part of one.

Another interesting part of the UUnet story leads me to my second "service based on free software" example. When Rick converted UUnet from a nonprofit to a for-profit corporation, he had to roll over the accumulated proceeds into another nonprofit. He funded an organization called the Internet Software Consortium, whose mission was to maintain free software required by the Usenet and internet infrastructure. Two of the packages supported by ISC are INN, the current generation of Usenet news server, and Paul Vixie's BIND (Berkeley Internet Name Daemon), the software behind the Internet Domain Name System.

The DNS also has turned out to have a huge service business associated with it, though that business was hijacked by a government contractor and turned into a lucrative monopoly. I'm talking, of course, about Network Solutions, whose business of registering domain names was a service business implicit in the DNS.

In a similar way, you can make the case that companies like Hotmail and Critical Path are exploiting service opportunities afforded by the net's open source email infrastructure.

What does this have to do with Linuxcare, and what lessons are there for entrepreneurs looking to capitalize on open source?

First off, service does not equal tech support! It does not equal training and certification. It does not equal custom development services! It can include all of these components, but it most go beyond them, by reframing a problem that your customers are struggling with, and offering a compelling outsourced solution.

Service means helping other people to get their jobs done, effectively becoming an invisible enabler of their business. When IBM or Oracle sells service as part of their software offering, they are really selling solutions, the assurance that some kind of business problem is going to go away, or some kind of business process is going to be carried on in a more streamlined fashion.

Successful service businesses don't necessarily create new revenue streams; they often divert existing ones.

Two companies that I think have a lot of promise in building exciting "service" businesses based on open source are Collab.Net and Digital Creations. (Disclaimer: I am an investor in both companies. But I am not high on these companies because I am an investor. I am an investor because I think they have got something right.)

Collab.Net's mission is to help manage open source and other internet-enabled collaborative development efforts. Just as UUnet realized that Usenet and email over UUCP required reliable communications infrastructure, Collab.net realizes that distributed internet-era software development projects (whether open source, closed, or somewhere in between), need tools and infrastructure for collaborative work.

Collab.net's first offering is sourceXchange, a marketplace for open source software RFPs. ("We need a Linux printer driver for our new HP printer, and we're willing to pay for it.") But that's just the tip of the iceberg. Collab.net is ultimately about bringing the best practices of successful open source software development projects to anyone who is developing software.

Collab.net founder and CTO Brian Behlendorf was a co-founder of the Apache project. Like Rick Adams, he didn't try to package and sell the software he'd worked on; instead, he set out to build business services around the tools, methodologies, and processes that have made Apache such a success.

Unlike many open source projects, which grew up around the work of a single visionary developer, Apache was from the outset a collaborative project. After the original development team left NCSA for various commercial web startups, the NCSA web server was orphaned until a group of its large end users decided to take on the job of maintaining and extending it by coordinating their patches. The result, "a patchy server", went on to become the dominant web server in the market, but the real secret of Apache may well be the development processes the team put in place. (For more information on these processes, see Brian's essay in the book Open Sources.)

Companies trying to harness open source often flounder because they think that all they have to do is change their license and put their code out on the net. They find that there's a huge amount of work required to develop consensus around a technical vision and an architecture that people can participate in, then to build momentum and coordinate community contributions. In the end, open source isn't just about new software licensing terms but also about a new, internet-enabled style of distributed software development.

While Collab.net's early work has largely been with vendors like Hewlett-Packard who want to cooperate with the open source community, the ultimate target market is far larger. As Eric Raymond noted in The Magic Cauldron, far more software is written for use than for resale. Any company that is spending money to develop software in support of its business processes is intrigued by the opportunity to harness programmers outside its walls, and to reuse source code that's already been debugged by a thousand eyes. Those companies that can figure out how to embrace the best practices of open source development and apply them to their own internal development efforts will be the first to recognize the breakthrough productivity gains that open source provides.

For example, Galactic Marketing is using sourceXchange to develop an open source workflow management system. Invisible Worlds is using sourceXchange to produce open source components that support its new XML-based Blocks data exchange protocol. The service that Collab.net provides to its customers is reliable access to the far-flung developers of the open source community. Much as UUnet provided internet dial tone, Collab.net provides "open source dial tone." (Or should we call it "collaborative development dial tone"?)

Digital Creations (the company behind the Zope application server platform) has a service-based business model for providing infrastructure for managing and building web content. They look at a company like Vignette, which gets a huge proportion of its revenue from web development services involved in deployment of its proprietary content management software, and believe that they can capture the same kind of revenue even when the software itself is a low or no-cost item. After all, platform software like Vignette or Zope isn't something that you can use all that effectively out of the box. You need to customize it heavily for your site, and that is often an outsourced activity. While anyone can download and install Zope, if you really want to take it to the max, who better to help you than its creators?

As Digital Creations CEO Paul Everitt said to me once in email, "We're awfully lucky. Digital Creations is in a market that, primarily due to the hiring crisis, wants solutions rather than software. At the same time, people pay based on value, and cost of software is rather immaterial. Obviously we're also lucky since perceived value with mainstream choices in the content management market right now involves six or seven digits."

Of the two companies, Digital Creations is closer to the pure professional services model now being explored by Linuxcare. But Digital Creations isn't ignoring the frontier of internet-enabled "e-services" either. They've just added support for XML-RPC to Zope.

I've written elsewhere about the way that things that used to be delivered as applications are increasingly being delivered as internet services. What's most exciting on the services front is the way that we're starting to see the development of protocols like XML-RPC and SOAP that will make it easier for companies to create electronic services that can be accessed by other web sites and remote programs. In the age of the net, "services" doesn't just mean professional services, it means figuring out how to build frameworks for automating activities that used to require human labor.

On that note, another open source company that seems to me to have a services vision at its heart is Eazel. (And no, I'm not an investor in Eazel, though not for lack of trying :-). The best I could do was to get founder Andy Hertzfeld (who was one of the key developers of the Macintosh) to keynote at our upcoming Open Source Software Convention in Monterey.) Eazel realizes that creating a next generation user interface for Linux means more than putting a pretty face on the system. From what I've gathered, Eazel's business plan includes not just a new desktop but services for software maintenance, update, and administration.

My point isn't to suggest that any particular approach to services based on open source is the best one. Right now, the open source sector is still young, with rapidly changing tools, practices, and services. What I hope is that the open source community will think big. The really huge opportunities typically redefine markets, creating services that seem obvious and necessary in retrospect, services that were once handled haphazardly but are now so central that no one can do without.

That's ultimately the lesson of UUnet and Network Solutions: to find the killer service, you have to stop taking things for granted. You have to look with fresh eyes at what problems people are struggling with, and find a way to take them off their hands. Once you've succeeded, everyone will take things for granted again.

Tim O'Reilly is the founder and CEO of O'Reilly Media, Inc., thought by many to be the best computer book publisher in the world. In addition to Foo Camps ("Friends of O'Reilly" Camps, which gave rise to the "un-conference" movement), O'Reilly Media also hosts conferences on technology topics, including the Web 2.0 Summit, the Web 2.0 Expo, the O'Reilly Open Source Convention, the Gov 2.0 Summit, and the Gov 2.0 Expo. Tim's blog, the O'Reilly Radar, "watches the alpha geeks" to determine emerging technology trends, and serves as a platform for advocacy about issues of importance to the technical community. Tim's long-term vision for his company is to change the world by spreading the knowledge of innovators. In addition to O'Reilly Media, Tim is a founder of Safari Books Online, a pioneering subscription service for accessing books online, and O'Reilly AlphaTech Ventures, an early-stage venture firm.

Discuss this article in the O'Reilly Network Forum.

Return to the Linux DevCenter.


Copyright © 2017 O'Reilly Media, Inc.