Mindset: Quality

Making software that works is easy. Making software that works most of the time is harder. Making software that works all the time is extremely hard, Getting to the "all the time" goal is almost an asymptotic curve, with time, cost, patience and perseverance on the other axis.

So this is more a mindset, the road to quality rather then a point at which quality can be achieved. 

Running a purely software company I often feel the pain others like us have been going through, Microsoft perhaps is the most obvious example. 

PC hardware is much better now and test / qualifying procedures are much stricter now. But roll back to early 1990's and some of you may recall hard drive controllers that worked with only certain motherboards, interrupt conflicts etc. This chaotic hardware landscape was the main reason behind the bad press Windows was getting, especially in comparison with Apple. Apple was lucky to have full control of the hardware. Or I should not say lucky - it was just their decision. Microsoft on the other hand relied on 3rd party hardware manufacturers.

And even the best software efforts can be ruined by poor hardware. And hardware - I do not mean only the silicon, as it very often includes some low level software, and this is probably where most often the "hardware" companies fail to deliver. Hardware mindset is different than software mindset, instead of this asymptotic path to the ultimate quality, hardware has a strict point in time when it is "done" and "ships" and no further improvements are made after that date.

Hardware vendors push towards reaching that milestone, very often aware of some existing issues, they take a calculated risk approach. Unfortunately such approach very often ruins the software quality efforts.

I write this from my own recent experience of a joint project with a traditional hardware company. I was struggling with their "different" approach to quality. We were providing 95% of the software for their product and their job was to build the hardware to our specs, also providing the remaining 5% of low level hardware abstraction level (HAL) software. We were fortunate enough to have strict test procedures (they have been designed and continuously improved based on similar engagements) and I was surprised the hardware vendor was still disputing our test results and recommendations.

Yes the ripple voltage on the interrupt line is high, but has it really triggered an unwanted interrupt? We could add a capacitor to filter, but if it does not trigger, we feel this is unnecessary.

No, it did not trigger. Under our test conditions. But (for example) considering the component tolerances, temperature impact, component ageing, we felt like there was just not enough margin to be sure it would never trigger an unwanted interrupt. It took us several weeks and multiple conference calls to convince them to make this little change. 

This partner clearly had a different mindset. To them, our quality mindset was a mindset of diminishing returns. And they are clearly not alone. There are thousands of cheap, not - dependable, misbehaving products on the market today. All could be characterized as crap. Why do we make them at all? What is the point of rushing out something that is a bit cheaper and will eventually fail to capture the market share because people will simply get frustrated with the brand.

Cheap things are very expensive for the civilization as a whole. In the end they are waste, polluting the environment and leaving this added residue of frustration among users who in the end will prefer not to invest in some technology at all.