Tells an entertaining parable about a conversation with a civil engineer who insists that bridges fall, period, and everyone knows it. Then turns to the timeworn (but understandable) observation that software engineering standards are absurdly weak compared to civil engineering standards.
It's all true! And should be taken seriously for mission-critical jobs (that's why I, a fan of Microsoft, would still not trust Windows running my autonomous vehicle).
But one can't help observe that their spiffy website could not possibly exist in this timeframe if consumer computer software had to be verifiably safe. It would be grotesquely behind where it stands today--people love features more than fixed software. We'd still be running browsers using, what, 1990 technology?
Even if it could exist in this timeframe, how much would it cost?
The issue I have with their viewpoint is that there doesn't seem to be any baseline for "good enough" that is short of absolute perfection forever (assuming adversaries that will attack with perpetually increasing sophistication.)
Alternatively, if the software we had was all well written, developing new software and bringing on the next generation would be so much easier and faster. The slowest development I have to do is on software that's horribly badly written; the fastest I do is on the most solid, most secure, most documented software; that's the software which I can and do improve and advance. The badly written junk is so much harder to advance.
If we did it properly, I suspect we wouldn't be on 1990 technology; we'd be using software that we won't actually see for another 20 years.
4 comments
[ 1.8 ms ] story [ 16.6 ms ] threadIt's all true! And should be taken seriously for mission-critical jobs (that's why I, a fan of Microsoft, would still not trust Windows running my autonomous vehicle).
But one can't help observe that their spiffy website could not possibly exist in this timeframe if consumer computer software had to be verifiably safe. It would be grotesquely behind where it stands today--people love features more than fixed software. We'd still be running browsers using, what, 1990 technology?
The issue I have with their viewpoint is that there doesn't seem to be any baseline for "good enough" that is short of absolute perfection forever (assuming adversaries that will attack with perpetually increasing sophistication.)
If we did it properly, I suspect we wouldn't be on 1990 technology; we'd be using software that we won't actually see for another 20 years.