Original URL: http://www.theregister.co.uk/2012/11/26/os2_final_fail/

Where were the bullet holes on OS/2's corpse? Its head ... or foot?

Ex-IBM insider Dom Connor reveals what went wrong

By Dominic Connor

Posted in Operating Systems, 26th November 2012 11:02 GMT

Part two My last piece on OS/2 was in part a mea culpa, a history of my part in its downfall. However, I can't claim all the credit. In fact, if I'm honest, there were hundreds of reasons why OS/2 failed, and most of them had nothing to do with me. So, here are some of the real corkers.

Once upon a time, IBM made extraordinary money out of big iron; mainframes so large that the corporation boasted a German university ran its heating system as a side effect of its Big Blue computer.

Even “mini” computers were often the size of a large room and that gave IBM huge control. It was just accepted by businesses that kitting out an organisation with a serious amount of computing power involved a serious amount of cash. IBM found it could charge whatever it liked - and did so. Personal computers didn't fit that model.

Then a threat to IBM's big iron emerged: beefed-up PCs in the form of servers crept up on Big Blue, and so the mega-vendor decided to return fire by turning its own PCs into minis and mainframes. A new operating system was needed to achieve this. DOS had to go.

Microsoft, which joined forces with IBM to craft OS/2, loved the idea of moving upmarket; at this point it had almost no presence beyond the desktop, and Bill Gates was entirely honest when he called Windows and DOS “stepping stones to OS/2”. He left competitors DR-DOS and Unix out of the picture.

What about OS/2 server?

Back then an x86-powered server had more oomph than the IBM System/36 and 38 minis and even the AS/400, which came out roughly the same time as OS/2 at the end of the 1980s.

You could stuff your Intel-powered box with disks and memory, and there was no reason to buy it from IBM, run IBM software on it nor pay IBMers to fix it and integrate it with other systems. You may think this is not only a good thing but inevitable. But if you own 85 per cent of a very lucrative market you don’t see that particular inevitability as anything good.

An original member of the System/360 family announced in 1964, the Model 50 was the most powerful unit in the medium price range.

An era of IBM machines, printers, disks, people, cats and dogs living together

Countless businesses had IBM software running on IBM machines hooked up with IBM SNA connectivity to an IBM printer that used IBM paper. Only a complete fool would risk losing that stranglehold, yet IBM faced losing control over the inherently disruptive and diverse world of PCs.

IBM flogged personal computers without any thought at all for connecting them to the rest of its systems. Indeed for a long time you had to get third-party hardware and software to achieve this. IBM begrudged every penny of that lost revenue, but it did notice that every PC attached to a mainframe became a terminal that only ever added to the big iron machine's processor load. That meant customers had to upgrade their IBM mainframes to keep up. The rule of thumb was that every smart terminal was worth half a MIP in upgrade sales.

I don’t know if that’s entirely true, but it was part of the justification for keeping PCs as PCs and not endorsing them as shared machines. Yes, IBM’s power was so great that if it did something then it was accepted as wise. There was, of course, Unix and VMS, but they weren’t from IBM and that scared many customers away.

An IBM x86 server platform, though, would replace juicy mainframe margins with the pathetic ones in the PC domain. That meant although OS/2 was structurally capable of being a first-rate server operating system, this superpower was not exploited until it was too late - and IBMs servers sucked hard on price.

IBMers in the corporation's PC division believed servers were going to be huge, but it was tacitly accepted that Microsoft would take OS/2 to servers. This avoided upsetting the parts of IBM that made real money unlike IBM’s PC division, which had a high profile but low profitability.

It’s hard to see how they could have made any other decision, not least because IBM was run by “business people”, defined as those who would regard any requirement to understand their products as offensive and got aggressive if more than a fleeting mention of them was made. Its competitors were run by people who had personally built some of their tech; you may not think Bill Gates was a star programmer, but he had written code rather than worked his way up from accounts. He ran rings around IBM.

The Red Mark of Death

As you may recall from part one of this series, I was a bug-hunting contractor working for IBM on OS/2 in the late 1980s.

It’s obviously been a while, but the passage below is what I remember of the wording on a blood red sticker that IBM’s tenacious security people would place on your desk if they found you'd left anything out at night that could be top secret.

By this mark I cast you sinner. Thou hast sinned. Yes, sinned in the eyes of God, sinned in the eyes of man, sinned in your own heart. My eyes can no more bear the pain of seeing your sin. Begone from this realm of grace and attend at the portal of the keeper of iniquities where your sins may be made plain to you.

Or something like that, anyway. Since the security sniffers couldn’t read disks, any they found were assumed to be full of secrets so they took them to a safe place and you had to grovel to get your disks back. To twist the knife you had points knocked off the evaluation scores of both you and your boss.

We’re alone here, so you can admit you really want one of those stickers, don’t you?

So did the Microsofties when they worked, depending on your view, with or in spite of IBM. After the first one got a sticker for leaving a programming manual on a desk, they all wanted one to take home and show to their friends.

Remember, Microsoft was not the dinosaur it is now. It was the cheeky small mammal eating the dino's eggs. It was a time when Redmond executives used their own products without resenting the time it subtracted from reducing their golf handicap. Yes, it was a long time ago.

Whoa, look out! It's crazy OS/2 in the house

These days I have a sideline in being an expert witness in IT-related cases, and even my own clients are extremely sensitive about letting me have access to their systems. That’s how IBM treated Microsoft even though OS/2 was supposed to be a joint effort.

IBM flatly refused to allow Microsoft to use any part of its network, which given the way Microsoft had behaved was entirely rational, even if it wrecked productivity. So the Microsofties created their own network using cables they got from the Tottenham Court Road shop that also supplied their PCs.

We’ve all had the joy of turning up at a new job and waiting for a working computer, but if you were the biggest IT firm on the planet and your most important project was slipping, wouldn't you make sure your partner's developers have PCs ready and waiting for them?

When Microsoft's finest pitched up in London, they bought their own gear. It was easier and they preferred Compaq machines. Unfortunately, the IBM PS/2, a desktop that was supposed to primarily run OS/2, was not fully compatible with the Compaq computers and potentially the code the Microsofties were rattling out. Microsoft in the US had to battle to get PS/2s from IBM to test the software.

Where was the database?

At the time IBM also owned the database market to the extent that some IT pros would struggle to name a competitor and probably just blurt out the name of an IBM product anyway.

Begrudgingly Big Blue eventually came out with an “Extended Edition” of OS/2 that could connect into IBM database systems, but it was pretty hard going to say the least. It was not until I used Lotus Notes that that I saw anything as bad. Certainly it was not seen by anyone as part of IBM’s well-respected suite of database management systems.

IBM made more money out of its database sales than the whole PC division put together, and it vetoed any attempt to bring out a database that was more user friendly than a cornered rat on a house of cards. I don’t actually know who wrote it, though I do know he has never been to England because he still lives, since I had write code for it and it hurt.

Decades later RIM made the same mistake: it made a tablet specifically incapable of doing email, which its BlackBerry smartphone did so well. It ended badly. RIM wanted to avoid "confusing the market", but really it's evident the smartphone team's management was politically more powerful than the newbie tablet gang.

Sign this, it says: 'Do not ever talk to our programmers again'

We learned at one point that distributing DOS-with-Windows cost less than just DOS, so Microsoft effectively paid hardware vendors to ship it. IBM salespeople weren't at all incentivised to sell OS/2, but at one presentation the guy handling IBM's largest banking client cornered me in a kitchen and offered to commit an act that would get both of us fired.

He wanted to speak to someone in the operating system's development team.

Apparently he'd had earlier promised in writing to not talk to anyone directly involved in producing IBM's most important software product. His own PC couldn't run it, but he was far more committed to IBM than it deserved and he really wanted to sell it.

I now teach bankers financial programming, but back then I was pathetically ignorant of software economics and finance. As was IBM.

OS/2 was priced by taking the development cost, dividing by the expected sales and adding a margin, which meant it cost hundreds. This was tough when Microsoft was paying people to take Windows. OS/2 also needed a couple of extra megabytes of RAM, which was a big deal, but worse was the fact that the only PCs it came on were from IBM, which was losing market share at a rate of knots.

Microsoft was ambivalent towards OS/2 by this time and IBM didn’t want non-IBM kit running its software, so no effort was made to persuade other manufacturers to ship the operating system with their computers. Installing OS/2 wasn’t very hard, but you couldn’t guarantee it would work on your hardware, and at about a thousand bucks a go this was an expensive experiment.

I've never worked for the London Stock Exchange - this will puzzle some who saw me tapping away at a screen there - but its main data terminal Topic 2 was going to be based upon OS/2. I appointed myself as the exchange's support guy because IBM wouldn't help anyone with OS/2 unless it was part of a mainframe project, and this wasn't.

This rule against supporting OS/2 even applied to IBM's top customers. The IBM rep for a client, which at the time was more lucrative than the US government, only got the support he needed when he risked his job by directly tackling people in the development team who realised that he was critical to our project’s success.

I got hate mail from managers for being among those who responded to the rep's pleas for help, and was one reasons why they eventually decided that my contract shouldn’t be extended.

Listen to your customers

One simple stupid mistake killed OS/2. It was called OS/2 1.3*. We'd got to the point where a 32-bit version that could run multiple copies of Windows with all its apps was nearly ready and it was a killer - basically delivering what you got in Windows NT but better and two years earlier.

Loading dynamically linked libraries wasn't done very well in version 1.0 nor 1.1, but some genius had worked out a much faster way for 1.3. Plus there were a whole pile of little fixes and improvements.

Then some customers told IBM they just wanted version 1.1 to run a bit faster and to use less memory. So development on the cool new version stopped and we squandered months on this pointless bit of crap.

At that point my contract ran out and I took to wandering the world doing OS/2 for banks and other corporates and, for self interest, took on the job of undoing some of the harm inflicted by IBM PR. This included installing OS/2 at Windows Magazine. IBM had flatly refused to give the mag a copy even though it was to be a favourable cover story because Microsoft had really pissed off the editor at the time.

Drivers, drivers everywhere, but not a byte to link

As a nearly competent test director at PC Magazine, I found myself trying to get the bloody thing to work on the random hardware we had lying around for a seriously big shootout between OS/2 and Windows. In those days PC Mag was a hugely important rag, so this really mattered. Windows just worked on pretty much every bit of hardware, OS/2 worked on very few machines and then only after serious force was applied.

This wasn't by accident. IBM actively tried to stop people writing OS/2 device drivers. The "sample" drivers were ridiculously flawed; I was present at one argument where the professionalism of one IBMer overcame his desire to keep his job and he flatly refused to let it ship because it was no good. He was overruled.

Booting up - and pray it gets further than this on your PC

I'm an OS/2 developer who learned Windows, so the benchmarks I wrote could not have been fairer to OS/2. In fact I knew quasi-documented software interfaces that were faster for plotting graphics although it wasn't particularly pretty.

The art director called the screenshots of the GUI benchmarks "the ugliest thing even seen on a computer screen". But my trick didn't work on most PCs and app developers were stampeding towards Windows, so with a lead lump in my stomach I cast my lot for Windows in the Editors' Choice.

IBM's PR department, which had never liked me, was really cold after that, but there was no further retaliation because it was already doing its best to make my life awkward.

IBM just didn't get PR, and the culture of secrecy meant its spinners saw the IT press much the same way the Royal Family sees the paparazzi. When it learned I had joined PC Magazine, an email was sent widely around Big Blue forbidding people to talk to "a journalist who you may have worked with in the past and know on a personal basis". I know this because I never knew how many friends I had at IBM until they all forwarded the email to me. I had many people from IBM at my wedding, including my best man and a former intern who was the bride.

For most of us, the fights weren’t just between Microsoft and IBM. Often they were technical staff versus managers, who repeated the mantra “we are not a computer company, we are a business” so often that within a few years IBM posted the largest loss ever made by a company in a single year since the “business” people produced almost nothing you might want to buy.

The emphasis on cost saving over producing a superior product was scary to watch. Another mantra was “every dollar saved on the cost of a PC is worth millions”, quite literally on posters all over the building.

The aftermath

Millions of copies of OS/2 were sold, running many whole firms, a large percentage of the cash machines and checkouts in the world, and I guess it made a profit on the cost of development. It lives on today as eComStation.

The project consumed resources that may have developed IBM a product that held its dominant position for longer. It also damaged the intensely valuable “no one ever got fired for buying IBM” maxim.

IBM now does far fewer products, more often buying in “mature” firms and using financial engineering to monetize them effectively, and of course is one of the top body shops. ®

Dominic Connor still is a member of the London OS/2 Drinking Group.

Bootnotes

* Why wasn't it called OS/3? When the name OS/2 was decided, no one had thought to check that OS/3 was available: it's a trademark owned by Unisys, not the sort of mistake you’d expect from a competent marketing group is it?

Also EMEA marketing was in thrall to the French who managed to stall the release of the UK English version until there was a French “European” build available; the delay cost us real sales.

The French version was harder to produce than you might think since all the dialog boxes and window layouts had been carved by hand in .RC files. I was one of the fools who didn’t realise that English is the most terse of all major languages. So during the internationalisation effort, it became obvious that the English-friendly dialog boxes either looked ugly or the text didn’t fit at all with other languages, which meant long hours at the end of the project manually aligning everything.