If you don't know much about hi-fi or home cinema, you can buy a complete system from one company. If you are a bit of a buff, you can build your own by choosing the best amplifier, CD player, speakers and so on from different companies and fitting them all together. Over the past two decades, the computer industry has gone through the same sort of process but on a much bigger scale, giving users an extraordinary range of choice. Now, not only can you buy PCs and servers from numerous different companies - IBM, Dell, HP, Toshiba, etc - you can choose between different processors, hard drives, graphics cards, and other internal parts.
In the beginning, like companies in many nascent industries, computer manufacturers were "vertically integrated". In layman's terms, they did everything. They made the hardware, wrote the operating system, languages and applications, published the manuals, sent out the salesmen, and ran the training courses. When it all went pear-shaped, they came round and fixed it. IBM, which had about 70% of the market, even runs a bank - IBM Credit Corp - to help you pay for it all.
This is only what you would expect. If you want to sell things into a new market, the best way to do it is to deliver "the whole product". In other words, solve a customer's business problem, such as operating a warehouse. If you deliver a couple of boxes and tell customers to hunt around for the missing bits, you are not going to shift a lot of kit.
The whole-product approach led to systems such as IBM mainframes, DEC minicomputers and Wang word processing systems where all the parts worked together, though, being proprietary, they didn't necessarily work with anything else.
But building "the whole product" is, unfortunately, expensive. Developing the IBM System 360 series of mainframes - on which western civilization still largely depends - cost more than the Manhattan project, which built the atom bomb. According to IBM's chief executive, Tom Watson Jr, the company hired 60,000 extra staff, invested about $750m (£456m) on engineering, and spent $4.5bn building five new plants to create the S/360. And these were 1960s dollars, so multiply by six to get the equivalent today.
Vertical integration had advantages for customers: apart from anything else, you knew where the buck stopped. It also had huge advantages for the suppliers, who could generally ensure that the buck stopped in their own pockets. Not for nothing were corporate IT departments called "IBM shops". They also got a wonderful advantage, known as "proprietary lock in". The customer was virtually a prisoner. As with the old Soviet Union, the only way out was to defect to another system.
According to "the whole product" idea, the advantage undoubtedly lies with the market-leading system, whatever it may be. This is a generally a good thing, because it provides stability, and there are huge communal benefits from standardisation. It creates what people commonly call "monopolies", but these can be short-lived because customers will switch when the benefits are bigger than the cost of switching. The ubiquity of vinyl LPs didn't stop people from switching to CDs. Nintendo's NES monopoly didn't stop the Sega Genesis/MegaDrive.
Indeed, if the switching costs are low, as with word processing software or video games consoles, you get a series of dominant systems - Wang, WordStar, WordPerfect, Microsoft Word; Atari, Nintendo, Sega, Sony. If the switching costs are eye-wateringly large, the same system sticks around for decades, such as the IBM S/360 mainframe.
But as an industry gets larger and more complex, more specialised suppliers start to thrive. This is exciting for customers, because suddenly there is the chance to break the suppliers' proprietary lock-in. Larry Ellison's co-founding of Oracle (as Relational Software) in 1977 is a typical example.
Oracle decided to create database software based on SQL, IBM's Structured Query Language, to run on different manufacturers' hardware. An IBM customer might be locked into IBM's DB2 database and a DEC customer into DEC's Rdb, but they could both use Oracle. This raised the tempting (if not necessarily realistic) prospect of swapping very expensive IBM hardware for less expensive DEC hardware or really cheap Intel-based hardware.
The key thing about Oracle's concept was that it offered a "horizontal" product, not a vertical one. Oracle didn't try to sell you its proprietary mainframe or network or word processor. You already had those. In fact, you probably already had a database, and knew how to use it. But rather than changing the whole vertical "stack" (DEC VAX mini, VMS operating system, DECnet network, Rdb, DECwrite etc), you could change just one horizontal slice. (I'm oversimplifying but you get the point. Andy Grove's book, Only the Paranoid Survive, has more detail.)
This approach has huge advantages for the customer, because choice creates competition, and competition drives down prices. And in the 1970s and 1980s, as more and more horizontal suppliers arrived, customers were able to choose "best of breed" products for each layer.
The personal computer industry went through much the same sort of development, but with the benefit of independent supplies of parts. (Apple did not have to invent the disk drive, just phone up and order some.) Still, many early micro manufacturers were vertically integrated and attempted to offer something like "the whole product" - Acorn, Apple, Amiga, Amstrad, Atari, Commodore, Tandy etc. Britain's tiny Acorn, with its usual hubris, not only created its own hardware, operating system, network and applications, it even developed a microprocessor - which became the enormously successful ARM chip.
But the "horizontal" approach was visible from the start, with hundreds of companies using Digital Research's CP/M operating system, microprocessors from Intel and Zilog, and the MITS Altair's S-100 bus. CP/M machines were considered clunky and boring by home users but businesses liked them: they were cheap and they did the job.
The horizontal approach really took off after the arrival of the IBM PC in 1981. Instead of building yet another proprietary machine, IBM, in a hurry, used parts from a couple of small suppliers called Microsoft and Intel. At the time, Microsoft had just 40 staff and a turnover that would not have filled IBM's stationery cupboards, let alone built an atom bomb.
IBM's clever if derivative design - basically it combined the ideas from the market-leading Apple II with those from the CP/M market - solid engineering, user-friendly "Charlie Chaplin" advertising and monopoly of large corporate computing made the IBM PC an unprecedented hit. It also opened the doors to thousands of rivals, such as Compaq and Dell. They bought the same parts from Microsoft and Intel and produced "clones" to run software written for IBM's machine.
With IBM setting a clonable hardware standard, customers soon had a choice at every layer of the horizontal stack. There were pitched battles between suppliers of compatible processors, operating systems, networks, word processors, spreadsheets, databases, printers, monitors and most of the things we take for granted today. And as cut-throat competition rapidly made PC-based systems cheaper and more powerful, almost all the vertical, proprietary suppliers were wiped out.
As Linus Torvalds, originator of the Linux operating system kernel, says in his book, Just For Fun, the opening up of the IBM PC "did more than anything to spur the PC revolution, which has in turn spurred the information revolution, internet revolution, new economy - whatever it is they're calling the massive changes now taking place throughout the world. It's the best illustration of the limitless benefits to be derived from the open source philosophy."
Look at it this way: you may be stuck with the heirs of the proprietary IBM S/360 for a few more decades, but 600 million PCs could be switched to GNU/Linux in about an hour.
But there ain't no such thing as a free lunch. The horizontal layering of the industry creates its own problems, just like assembling a hi-fi from separates. If anybody can build and sell a computer that can run any software, there is very little quality control, and millions of people end up with combinations of hardware and software that have never been imagined before, let alone tested. Rather than being spoon-fed by high priced vertical suppliers, customers have to become their own systems integrators and IT managers. The products may be cheap, but freedom has its price.
