The great thing about the computer industry is that everybody believes in standards - especially if they own one. If they don't, at least they can get together with a group of rivals and agree one. This is always met with widespread praise - which is nice, because this kind of activity might otherwise be considered illegal collusion to distort the market and avoid competition.
However, governments are usually happy about standards agreements because the alternative is incompatibility, which is always harmful for consumers. The problem is that everybody wants competition, and everybody wants standards, but you can't always have both.
You can, of course, have a competitive market based on any number of products designed to be mutually incompatible. The games console business is an example. That's why most families have at least one obsolete games console stashed away somewhere, and are wary of buying another.
These battles almost always result in one company taking the majority of the market, and a string of barely remembered failures. Not only does this result in financial losses for the losers, es embody a vast waste of resources. Third parties cope with this predictable result either by avoiding minority markets or charging higher prices. Incompatibility is inherently expensive.
You can also have incompatible systems co-existing. This happened in the American mobile phone market, where it delayed the take-up of mobile phones. Europe gained a lead because we agreed a common multinational standard, GSM. This also helped the growth of local manufacturers, such as Nokia and Ericsson: the use of a European standard in effect protected them from Japanese competition.
But typically, incompatibility has a chilling effect on markets. Buyers delay their purchases until one winner has emerged from the bunch of losers. Thus DVD has been a success because the various manufacturers negotiated a common standard in back rooms, instead of having a Beta v VHS-style battle in the marketplace. But the rewritable DVD and DVD audio formats have not been a success, because consumers are still waiting for a resolution to the incompatibility of rival systems.
In an ideal world, all standards would be open, free, and tested. They would also work properly. This rarely happens. Such standards take too long to develop: committee meetings are held months apart across three continents. The standard tends to be overspecified, because market needs are uncertain in the early market, and because every supplier wants to have its own contribution included - especially if there's a royalty in it. Finally, products designed around complex and cumbersome standards tend to perform badly. There's an old joke about an elephant being a mouse specified by a committee, and OSI networking was an example.
The alternative is to let the markets define standards, then ratify them later. For this reason, most people recognise two types of standard: "de jure" standards (suppliers voting on committees), and "de facto" standards (users voting with their wallets). De facto standards do wonders for the profitability of the winners. They also seem to be a source of despair and resentment for the people who backed the losers, and spend the next 100 years whining about it.
With an "open" standard, the specification is published, so that any company can make a compliant product. Instead of choosing between competing products or ideas, users get to choose the best implementations of the same idea. With a "free" standard, any company can make a compliant product without paying fees for patents and other intellectual property rights. Since not many companies want to give away their intellectual property, however, standards bodies often settle for what is known as Rand: "reasonable and non-discriminatory terms".
One of the problems with designing products to published specifications is that almost every supplier finds a different way to interpret them, so the different products don't actually work together. They are de jure standard, but useless. The industry solution is to have regular "plug fests" - meetings where all the companies get together to test their interoperability. Products that work together can then be given a trademarked logo, which is a big help to consumers. For example, if you buy a wireless card for your notebook computer, it is much more important to have the Wi-Fi logo (which means it must pass the interoperability tests) than an assurance that it meets the published IEEE 802.11b standard. With de facto standards, however, compatibility and interoperability are the only things that matter. If you can "reverse engineer" or emulate your rival's product so yours works identically, no one cares whether it meets a written spec or not.
The customer benefits of compatibility and interoperability are obvious. However, the computer industry had to learn this the hard way, as usual.
If you have been following this series, you will remember that the computer industry began with vertically integrated companies. Each supplier tried, as far as possible, to deliver "the whole product" that met customers' needs. (For a discussion of "the whole product," see Schofield on Saturday, January 25.) This could well entail developing its own hardware, operating system, applications and, naturally, its own networking.
IBM, the computer industry's largest company, had its own proprietary way of connecting computers called SNA (Systems Network Architecture). Digital Equipment Corporation (DEC), which led the minicomputer business, had DECnet. Some personal computer companies did the same: Apple had AppleTalk, Corvus had OmniNet, Acorn had Econet. Proprietary networking added to the "lock in" that prevented users from exploiting a wide range of commodity hardware and software, and made it expensive to switch to rival systems.
The growth of the PC industry, and the arrival of "cross platform" software such as Unix and Oracle, tipped the computer industry into a horizontal form of organisation. The idea was that instead of using a proprietary supplier's vertically integrated "stack," users could choose a different product at each level. Instead of running, say, IBM's DB2 database on an IBM mainframe and DEC's Rdb Database on a DEC mini, you could run Oracle on both. (See Schofield on Saturday, February 1.)
In the PC business, the leading cross-platform networking system was Novell Netware. Novell's goal was to have Netware run on every significant platform - DOS, OS/2, Macintosh, Unix and DEC's VMS - and to interoperate with the leading networks, including ARCnet, IBM's favoured Token Ring, and Ethernet.
But at the time, western governments were obsessed with the idea of open systems. They were frightened that IBM's market power would make SNA the de facto standard, so they encouraged the development of a similarly heavyweight but open system called OSI (Open Systems Interconnect). This seven-layer model was adopted by the ISO (International Standards Organisation), and the US, UK and other governments tried to make its use mandatory within government. The hope was that the Gosip (Government OSI Profile) market would be big enough to attract suppliers and enable them to take over the commercial market as well. The whole sad story should be engraved on the eyeballs of anyone who thinks government interference is a really good idea.
However, the computer industry (and therefore the world) got lucky, not once, but twice. The first bit of luck was Ethernet, which is used for connecting computers together on a local basis. Ethernet was developed by Dr Robert Metcalfe at Xerox Parc (Palo Alto Research Center) in California, and therefore did not get weighed down with multiple kitchen sinks inserted by all the people on standards committees who work for competing suppliers. Since Xerox didn't have the clout to get Ethernet adopted, it formed a multi-vendor consortium and, in 1980, published it as the DIX (DEC/Intel/Xerox) standard. The technology was then adopted and ratified by the Institute of Electrical and Electronics Engineers, and became IEEE 802.3 in 1985.
Today, Ethernet is becoming popular running over the "ether" in its Wi-Fi wireless format, which is technically IEEE 802.11b. Ethernet has become a template for how you get a standard established. Take a great idea that actually works, which usually means it was developed by one person or a small team. Launch it with multi-vendor backing, preferably from a consortium. Publish it to make it widely available. Get it ratified by an independent standards body. Sadly, this isn't a great way to get rich. However, Metcalfe founded 3Com (for computers, communications and compatibility), which did make a lot of money selling Ethernet networking (see Bob Metcalfe interview).
The second bit of luck was the internet protocol, TCP/IP, which was designed to connect computers together on a global basis. TCP/IP (Transmission Control Protocol/Internet Protocol) came out of the development of a university research network, which was called Arpanet after Darpa, the US government's Defense Advanced Research Projects Agency. TCP/IP was developed in the 1970s by Vint Cerf and Bob Kahn, and standardised in 1978. Versions were created for all the important operating systems, and in 1983, the network was switched over from the Arpanet's NCP (Network Control Protocol) to the Internet's TCP/IP. Nobody ever thought IP was going to become a commercial success, but when the commercial world needed it, there it was.
TCP/IP shows that important standards can be developed in a pre-competitive environment, rather than driven by commercial considerations. Nor is this the only time it has happened: for example, in the 1980s, the X Window cross-platform windowing system was developed under Project Athena at MIT, in collaboration with DEC. But I would not bet money on it happening often.
Generally, standards come either from market victories or are the result of horse-trading among the companies that would rather not take an "all or nothing" bet on doing their own thing. To that extent, a standard is usually an attempt to rig the market. However, since IT markets depend absolutely on compatibility and interoperability, consumers usually benefit, and it leaves a lot less blood on the floor.
