| June LJ Supplement |
Linux Journal: So how do you propose that enterprises make sense of Linux' appeal? Because clearly there is some appeal, and it is hardly limited to Web servers at ISPs.
Craig Burton: What we need is a conceptual framework by which enterprises can start to understand a larger world that includes Linux as well as NT, open source as well as closed source, and problems that neither has yet solved – and there are many. Fundamentally, there are two issues at hand here. One is interoperability. Problems with interoperability transcend the virtues and liabilities of individual operating systems, applications and the protocols that should – and too often don't – allow them to work together. And the simple fact is that Linux and the applications it hosts will not be adopted by enterprises unless they interoperate with those enterprises' incumbent operating systems and applications. The other issue is business model. Open source development sounds very promising. But let's be real here: it remains unproven for the kind of goods that Microsoft and its closed source competitors provide in abundance and customers trust.
Linux Journal: What do you mean by a conceptual framework?
Craig Burton: At The Burton Group, we developed one called the Network Services Model, back in 1991. All of the major vendors have since saluted it, and most of them talk about network services in terms we provided for the first time with that model. But practicing it is another matter. Doing what's right for the customer, and for the marketplace, is not easy. And I submit that it's no easier for open source development than it is for closed. Everybody has some kind of agenda or subagenda that can be unfriendly to the market. With most commercial developers, it's trying to lock in customers and lock out competition. With open source developers, it's the open source religion. Tying to knock off Microsoft is not necessarily a way of serving the marketplace. It's a way of serving egos.
Linux Journal: Can you tell us what the Network Services Model is, in brief terms?
Craig Burton: Sure. It describes the basic rules by which different goods from different vendors should compete and cooperate in a networked world. It describes a set of services required of networks – notably file, print, management, security, directory, messaging and Web. And it predicts that these services will only scale and proliferate to the degree they are interoperable. It also says no one vendor or standards group can to do it all. That said, there are five elements to the model:
Linux Journal: So how do you define or qualify a network service like security or directory, especially when different vendors and standards bodies have different definitions and agendas?
Craig Burton: That's a good question, because it speaks directly to what customers want to see from Linux and the open source community behind it. To qualify, a network service must meet the following technical guidelines. It must have:
Here's the kicker: all four must be well defined, documented, and functional. Are they for open source?
Linux Journal: Well, there are a number of standards bodies in and around the open source movement –
Craig Burton: – Hold it right there. There are three kinds of standards, and two of them are to a high degree outside the scope of standards bodies. Those are:
De jure standards are "official" to the degree they been sanctioned by a standards body. The Internet suite of protocols is a good example of a de jure standard. De facto standards are defined by the market rather than by a standards body. Most legacy applications, services, protocols, and tools are all de facto-based products. Windows, Win32, NetWare, IPX/SPX, and SNA are all de facto standards. De rigueur standards are those based on the habits of markets, rather than by their conscious decisions. Or, in French, "necessary according to convention." In other words, beliefs. De rigueur is the subconscious side of de facto. The de rigueur standards Linux confronts might go like this: "Linux may be good for Web service, but not for anything else." Or "open source is unaccountable."
Countless de rigeur standards have been spectacularly wrong. "The world is flat." "The world is the center of the universe." "Precise latitude is impossible to calculate without a viewable landmark." "Manned flight is impossible." "You'll never need more than 640k of memory in your PC." One de rigeur standard we need to change right now says "proprietary means closed." In fact, proprietary is just one among several choices. Here is how they look on a matrix:
The opposite of standard is not proprietary. A standard is something defined by de jure, de facto, or de rigueur practices. Something "proprietary" is something owned. If a company owns the intellectual property rights of a product implementation based on standards, it does not define the product as part of the public domain. It defines the product as proprietary; because that company owns the product.
In this scenario, proprietary is orthogonal to standard and the opposite of standard is simply non-standard. The intermittent windshield wiper — now a standard — was based on non-standard technology at the time of its invention. It wasn't defined by a committee; it wasn't de facto, and certainly not de rigueur. The car manufacturers took one look at it, loved it, showed the inventor the door and knocked off the invention. To make a long story short, the inventor finally got wealthy by proving ownership of a nonstandard technology. The intermittent windshield wiper is a perfect example of a proprietary de facto and de rigueur standard.
On the other hand, there other many examples of de jure standards that are not proprietary, but "public-domain," that have never been successful. The OSI network transport protocol is a good example. The US government "mandated" OSI compliance. It never happened, and never will. As Scott McNealy said, "Big hat, no cattle."
This next matrix is a map of reality. Whatever you do — and whatever your customer chooses — is going to fall somewhere in here. It may be your religion to live in the upper right or the lower right corners. Or it may be your calling to offer goods in the upper left. The strategic way to look at this is: what works best where, and when? Which standard do you want to use? What do you want to keep proprietary or put in the public domain, and what are the trade-offs? And remember that there will always be tradeoffs.
Notice that the opposite of Gratis License is License for Fee and Open Source is orthogonal to each of these terms. The opposite of Open Source is Binary Source.
Linux Journal: And there are hybrids.
Craig Burton: Right. Parts of the Netscape Navigator and Communicator are now in the Open Source/Gratis License quadrant. Additionally, there are elements of these products for which Netscape owns unlimited use as a result of License for Fee. However, the license agreement does not allow them to place the source in the Open Source/Gratis License domain.
Linux Journal: What about Windows?
Craig Burton: Windows is the classic Binary Source/License for Fee technology. Win32 is a classic example of straddling multiple quadrants. Microsoft's instinctive and masterful use of this matrix is at the root of Windows dominance.
Linux Journal: And developer dominance.
Craig Burton: Right. Microsoft freely gives away, and even pays, for vendors to use Win32 to develop client and server applications. The source to the Win32 implementation is proprietary, standard, and License for Fee. The Web succeeds because it is a classic example of technology and business practices that follow the guidelines and principles defined by the Network Services Model. Every single standard that makes up the Web – including TCP/IP, DNS, HTTP, HTML, DHCP, and on and on – is being pushed by the envelope of innovation to the very limits of possibility.
Linux Journal: How do the standards guys play in here?
Craig Burton: It is both painful and comical to watch the standards bodies – the IETF, W3, DMTF, whatever – grapple with the rate that innovation is pushing the standards process to its limits. Look at Tim Howes and Mark Smith's book on LDAP, which was published in 1997. The book was probably written from 1994 to 1996. At the time of publication, adoption of LDAP v3 was "imminent," they said. In those days it was generally agreed that LDAP v3 would be ratified with flying colors by mid-1997. Here we are in 1999 without any clear end to the debate. So innovation is everything. It is the fundamental market reality. Competition and functional needs are met by innovation. No standards process could possibly adhere to the pace of unhindered innovation.
Linux Journal: Now what about interoperability?
Craig Burton: Can we all get along? To the customer, interoperability means cheap. To the vendor, interoperability means potentially giving up market share to another vendor. To the software developer, the availability of interoperable products means a bigger market and potentially more revenue. Yet interoperability between multiple vendors' products is fundamental for freedom of choice, which is something markets always prefer. Vendors at best like to meet this imperative half-way: on the client side. Not the server.
In other words, it is currently in the best interest of vendors to encourage other developers to support their services and server applications by diligently defining the client side of interoperability. At the same time, they reserve the right to keep server- to-server interoperability undefined. In our first matrix, vendors are placing client-to-server interoperability technologies in the standard/proprietary and standard/public domain quadrants. There is even activity in the non-standard/proprietary quadrants. The game being played is to keep server-to-server interoperability technology very close and under wraps. This is why LDAP is a good client-to-server protocol, but is inadequate as a server-to-server protocol.
Linux Journal: What about service interoperability?
Craig Burton: Bottom line: build a server application without making each service separate and interoperable, and you're in trouble. Lotus Notes is a classic collaboration system that completely blurs the distinction of network services and a server application. In Notes, network services and network applications are jumbled together. There is no clear boundary between them. On the one hand, this is why Notes remains the most comprehensive system for building rapid deployment collaboration applications. Every other vendor is miles behind IBM/Lotus in providing a system capable of competing with the core functionality found in Notes. On the other hand, the Notes architecture is why IBM/Lotus has a major problem on its hands. Separating Notes services from Notes applications is no easy task. It's a rewrite. IBM/Lotus has to continue selling the integration of both services and server applications as a benefit, while it races to make its discrete services interoperable. Not easy.
Linux Journal: In Lotus' defense, Notes was ahead of its time.
Craig Burton: Exactly. Iris Associates, the designers of Notes, had no choice but to build their own infrastructure. Commonly available infrastructure for collaborative system development didn't exist pre-Web. Netscape is the classic example of a vendor that jumped into the fray of collaborative application systems infrastructure that was built on existing technologies. It was a great strategy.
Linux Journal: Their strategy was one that at the same time put them into a new race and invited competition in the race.
Craig Burton: Right. They went for services that were interoperable at both the client and server side of things. This also meant an opening for competition. When history gets written for this period, Netscape will be remembered for a lot more than just introducing and ubiquitizing the browser. Perhaps their biggest success was in helping define a bunch of services and then getting competition going in those markets by leading the way.
Linux Journal: Were they alone in understanding these principles?
Craig Burton: Close. But these concepts aren't too complicated. If you want to be a player in the technology race, it means understanding these principles and playing tough and fair. Worry about the customer need; foster the industry first, the competition second. Building on top of existing services where possible and being prepared to move from your own infrastructure, when appropriate, are imperative positions.
Linux Journal: Have we seen anything before that reminds you of Linux?
Craig Burton: Yeah. OS/2. By all accounts it was a much better platform than Windows. DOS, Windows and NetWare were the uncircumcised and the unclean in those days. But it lost. It was clear in the opinion of the customer that the long term safety in an OS/2 choice was really low. Once again, the defining issue wasn't the cost of the product, because IBM was really ready to give it away. (In fact, they couldn't give it away.)
But Linux has turned the corner as a strategic long-term platform. It has moved much farther than OS/2 ever did. What is causing that? That has to be explored. And you're not going to get the answer from the guys who want David to fight Goliath. Distingushing between OS/2 and NT, and looking for how Linux compares between those two, is a really good avenue to explore. One was successful. Another was not. Think about it. OS/2 had lots of stuff Linux doesn't have, especially money, and didn't succeed.
Linux Journal: Why did OS/2 fail?
Craig Burton: OS/2 failed because IBM and Microsoft together had a plan to take over the marketplace and remove choice from the customer, and the marketplace didn't like that. It was fully evident. OS/2 was very hardware specific: IBM used it to push its Microchannel. If it succeeded, only IBM could sell hardware. Microsoft used OS/2 to get rid of NetWare. It was a conspiracy to build technology to lock the customer in. Fortunately, the conspiracy was uncovered. The key here is that IBM and Microsoft were trying to solve their own agendas, and this was discovered by the customer base, who didn't like it.
Linux Journal: Which sounds nothing like what the Linux vendors are doing.
Craig Burton: Maybe not. But the Linux community has an agenda that's just as selfish and market-insensitive as Microsoft's and IBM's was with OS/2. It's an anti-Microsoft agenda. The Linux community wants to bring Microsoft down. Business customers may share some antipathy toward Microsoft, but they also like and depend on a lot of what Microsoft provides.
Linux Journal: What changed because of OS/2s failure? Did IBM and Microsoft learn anything?
Craig Burton: They got new agendas – ones that worked. Microsoft commoditized NetWare by emulating it, and IBM succeeded just by making better hardware.
So let me ask: what are the Linux vendors' agendas? To get the biggest market share of something they give away? That's an awkward position. Somehow Red Hat has pulled it off, though. (Or have they? They must be making money somehow.) Does Linux meet the business requirements for long-term success? That can only be answered by realizing that there is no single Linux, and it depends on the business policies of the individual suppliers of Linux to corporate customers. In the long term that will decide if this is successful or not. And my guess is that those suppliers are still too immature to answer that question yet.