/>
X
Home & Office

Is good enough good enough?

I was browsing my bookshelves the other day when I came across a book that I acquired (and I think I reviewed while working on PCW) over 20 years ago, called 'LANs Explained'. It was released in 1988.
Written by Manek Dubash, Contributor on

I was browsing my bookshelves the other day when I came across a book that I acquired (and I think I reviewed while working on PCW) over 20 years ago, called 'LANs Explained'. It was released in 1988.

When I re-read it, I was taken firstly by the way that networking technology has moved on, especially since networking seemed to go into a bit of a quiet period in the late 1980s and early 1990s, when little seemed to be moving in terms of speed and reliability. There were lots of ideas -- ATM or VG-AnyLAN anyone? -- and lots of standards but a good enough, do-it-all technology was missing.

But what shines from the book how important standards were perceived to be at the time. After all, without standards, how could networks interoperate? Those standards: the OSI seven-layer model was dominant in everyone's thinking, and what each layer did was highly prescriptive. In practice, of course, every vendor did it slightly differently -- and the result was frequently a mess, and generated a huge amount of finger-pointing. This layered concept is a highly useful way to think about networking, and I'd argue that the first four layers persist, broadly speaking, pretty much as the traditional model describes. Above that, it all starts to get a bit messy....

What the author, on W Scott Currie, had in mind, especially in his 'Future Trends' chapter, was a world where lots of different network technologies co-exist, and where they would all need to talk to each other. Hence the need for rigidly adhered-to standards. What Currie didn't envisage -- and to be fair, few others did either -- was a world where all the other standards just melted away.

Where is Token Ring today? Invented and backed by IBM, the world's biggest computer company, TR seemed at one point as if it would dominate. Yes, it was expensive and complex but it worked and was deterministic: you knew what bandwidth to expect and it delivered.

But quietly, the world was installing Ethernet. Once it was able to run over a reliable layer 1 infrastructure -- the hub-based star configuration -- it worked as long as you didn't overload it. Crucially, it was cheap to install. Then switches came along and Ethernet's biggest drawback -- packet storms created by collisions once the network load hit around 70 percent -- disappeared. This meant that Ethernet was good enough for almost every purpose.

The combination of Ethernet and TCP/IP now predominates. Neither is perfect, neither provides everything everyone would want from a networking infrastructure. But they are good enough and that's why they've been accepted as de facto standards. The rest have gone.

So when a technology vendor looks me in the eye and talks about how strongly they believe in adherence to standards, a voice whispers in my ear asking: but shouldn't it just be good enough?

I'd be glad to hear of examples of other technologies that have succeeded in this way -- and about those that you think might achieve greatness in a similar way.

Editorial standards