ISO never learns, IBM does

IBM slating ISO's standards work? That is possibly the most ironic It event of the year - but I'm afraid you need some ancient history to appreciate it.

Thirty years ago, the boot was on the other foot. ISO blasted IBM's proprietary practices. IBM has moved on since then, but it looks like ISO has gone backwards.

In the 1970s, IBM built up a lucrative proprietary business around the SNA network protocols it used for its mainframe systems. ISO was at the head of an effort which proposed standards to break this monopoly.

In 1977, ISO published the OSI (Open Systems Interconnection) seven layer network model - a fine expression of a basic principle behind open networks: write the protocols in layers.

Then it made a big mistake. It went ahead and wrote a set of "OSI standards" for communication protocols, which wasn't such a good idea. With the ISO imprimatur, these became government requirements, even before they were properly implemented. ISO wasn't a network engineer, and the protocols turned out to be unworkable.

They were never really used but, incredibly, it took more than ten years of heavy expensive promotion, before the OSI effort fell apart in 1996. As an entertainingly contentious Wikipedia article puts it, this "severely damaged the reputation and legitimacy of the organisations involved, especially ISO."

In the end, the old proprietary protocols died away. But they weren't replaced by ISO's creations. Instead, they were displaced by pragmatic implementations of truly open protocols, especially TCP/IP. OSI diehards couldn't accept other people's standards and tried bizarre rearguard actions against TCP/IP, but OSI was doomed.

IBM took the lesson well. It formally "adopted" OSI when Governments demanded it, but it learnt the deeper lesson and actually adopted TCP/IP, the stuff that worked.

ISO seemed to learn something. It learnt that standards come into standards bodies from outside. It learnt, more or less, to keep its head down. To my knowledge, TCP/IP has never been "ratified" by ISO. It just IS the international standard.

((Incidentally, around the same time, ISO adopted Posix, a programming interface for an operating system. It was developed elsewhere, based on working technology - Unix - and touted as the basis for an open standards environment which would be an alternative to the product from the would-be monopolist, Microsoft. Today, Linux, Unix and Mac systems comply with Posix, but it's not clear how often this is actually used).

Fast forward to OOXML. On one level, it's OSI all over again. Once more, ISO has published, and promoted for government use, a standard that - from many accounts - doesn't actually work very well. The ISO ballot resolution meeting, remember, didn't resolve all the issues, and Microsoft has admitted it will get the standardised version of ODF working sooner: Office 2007 will never support standardised OOXML.

But on another level, it's very different: ISO got this standard from the monopolist. It's as if, in the 1970s, ISO had standardised a broken version of IBM's SNA.

ISO clearly still doesn't get it. And this time round, IBM is trying to put it right.