> Fujitsu should take no blame at all. Their software is crap. But so is everyone's. That's software. We know and accept that in the industry.
Eh, it's one thing if a PC game lets you fall out the world. Such bugs are low-stakes, thorough checks aren't possible for performance reasons, and nobody denies that bugs are widespread.
But for an accounting system to lose track of money? Which is extremely high-stakes? And they've got more than enough time to use things like SQL databases with transactions and constraints? And they ended up telling the courts they were certain enough about the system to jail people based on it?
Can you name a single accounting system that hasn't lost money over the years? All the big players had major major flaws. I agree with the parent that although Fujitsu is at fault, theirs is honest: i personally find the criminal fault to be with managers and judges who will convict someone based on proofless software-produced "evidence".
If i as an individual who knew nothing about anything went to testify against 700 postmaster/postmistresses, would they ever be convicted? Why would a computer program have more credibility than me, a random bystander? Computers are just weird things manipulating bits, and noone should ever trust them for anything serious, especially to convict other people of wrongdoing.
My personal legal interpretation is that the judges failed (as they very often do) to uphold the legal standards of presumption of innocence and of examining the seriousness of claims produced in the tribunal. The fault rests with them.
EDIT: of course if fujitsu engineers testified that they would jail people based on data produced by their software, that makes them responsible. but the judge is equally responsible not to have independent experts (plural) verify those claims
> of course if fujitsu engineers testified that they would jail people based on data produced by their software, that makes them responsible...
The engineers would not have been testifying about sentencing, but about whether the system output was accurate. You yourself have just argued that all accounting systems have flaws, the engineers undoubtedly knew that, and if they testified to the accuracy of the output, then they would be culpable of making material claims that they knew had not been verified, and could be wrong.
> ...but the judge is equally responsible not to have independent experts (plural) verify those claims.
It is very obvious in retrospect that quite a few people should have been more suspicious, but that would do very little to absolve Fujitsu of the high moral and legal responsibility of telling the truth, the whole truth, and nothing but the truth if they were called upon to do so. "You should have checked what I said" is neither a legal nor a moral excuse for perjury.
Update:
"The data used in court has since been proved to be wrong and former Fujitsu members of staff are currently being investigated by the Metropolitan Police for potential perjury in the trials of subpostmasters that were blamed for unexplained losses."
Thanks for clarifications, that makes perfect sense. I'm still more angry at the criminals-in-robes we call judges, but of course testifying your software has zero flaws (in any case, and especially if you know otherwise) is ethically wrong.
Fujitsu's responsibility was to deliver code that met the requirements, on-budget. It was the Post Office's responsibility to ensure that the software was fit for purpose. And in fact they knew it wasn't; but they accepted it anyway, probably because having it re-worked was going to push them over-budget.
The PO's requirements for the counter-mech project consisted of three yards of requirement documents, in ring-binders. I spent several months in a room with no windows, working through those requirements. They were mainly forms, supposedly describing inputs, outputs, reports, and screens. MOST OF THEM WERE BLANK.
I wasn't with Fujitsu; this was an earlier iteration of the project. But I have no reason to believe that a step-change occurred in the way the PO handles IT projects, between when I worked on it and when Fujitsu picked it up.
> And they ended up telling the courts they were certain enough about the system to jail people based on it?
I think that should open them up to unlimited liability. How hard would it have been for them to audit a single problem site to find out their software was trash?
Sorry you're right. I don't mean to dump on my fellow coders. I've
actually done some safety critical stuff in my time and still have
nightmares 20 years later as to how its running in production.
I guess what I'm trying to say is that the distance between what we
understand as the probable correctness of code, and the fantasies of
regular folk in management or politics is enormous and widening.
The difference between our informed standards, and most uninformed
expectations, is dangerous.
> They ended up telling the courts they were certain enough about the system to jail people based on it?
Is that actually something people from Fujitsu testified to in open court? I could not find that in the original article, so I presume you have a different source? I would be interested in that source!
What standards? We accept more failure from software than pretty much any other product. If your car’s engine broke down as much as windows you’d never buy from that company again.
Crap software has trained us to accept complete random crashes, and if you look at most code written today you’ll realize it’s only getting worse.
Even accounting software can sometimes have bugs. Not saying Fujitsu doesn't deserve blame for making shitty software not up t standards and face repercussions, but 100% bug free is a pipe dream.
This is more about the organization literally driving people to suicide over believing that the software has a bug, but if they handled it correctly - i.e. not tell people that it couldn't be their software and the people must've stolen the money. - that isn't Fujitsu's fault.
Eh, it's one thing if a PC game lets you fall out the world. Such bugs are low-stakes, thorough checks aren't possible for performance reasons, and nobody denies that bugs are widespread.
But for an accounting system to lose track of money? Which is extremely high-stakes? And they've got more than enough time to use things like SQL databases with transactions and constraints? And they ended up telling the courts they were certain enough about the system to jail people based on it?
Our standards aren't that low.