• When a computer’s hardware goes wrong, it is highly unlikely to change the orderly execution of a computer’s instruction set that represents the software running. Typically, gross symptoms occur, for example the machine generating a lot of garbage symbols in its output, which will be patently obvious,
• When a computer’s software goes wrong in the sense that an unexpected outcome is delivered or the program aborts or “hangs”, for example, apart from the latter, there is no obvious way to know if the internal software logic has been wrongly programmed or some software procedure used to correct data or other errors has inadvertently created further errors.
Having to prove that computer hardware is functioning correctly is a ridiculous requirement.
Having to prove that software is working perfectly (i.e. exactly as specified and 100% bug free) is, in practice, impossible expect for the very simplest and often trivial programs. There used to be an 1970s endeavour called program proving, which would theoretically prove that commercial software was bug free. This was abandoned. It can work in very limited circumstances to show, for example, that a self-contained algorithm does function reliably and predictably. Program testing is the way that software is checked for bugs. This is normally done nowadays where test data is created to “stretch” the program and then the output is verified to check that it corresponds to the specified requirements. This cannot and does not eliminate all bugs.
To summarise simply: The probability of a computer hardware fault causing software instructions to malfunction is near zero. The probability of software containing bugs is near to 100%.
On the “footnote on recusal”: it is important to note a distinction which is not clearly spelt out in Conor Casey’s Policy Exchange paper to which you link but, reading between the lines, he clearly accepts - and is plainly right.
That distinction is between a general question (“has the AG advised on issues where he has a conflict”) and a specific question (“has the AG advised/been asked to advise on issue X, where there seems to be a conflict”).
The problem Casey skates round is that at least one of the questions that he cites is a specific question (Jenrick’s “has the AG recused himself from advising on any matter involving Gerry Adams?”).
If the AG answered that positively (“yes I have recused myself”) he would in effect be disclosing that the Law Officers have been asked for advice on the Adams matter.
That would breach the convention (rightly accepted by Casey, for the reasons he gives) that the Law Officers do not say whether they have advised on any particular matter.
Could he reply: "I can't say whether or not I have recused myself in any specific case but if I am (or have been) asked to advise on Gerry Adams then I would recuse myself"?
Yes I suppose that that’s right (although I am not sure as a matter of fact whether he would be precluded from ever advising on anything to do with Adams just because he acted for him on one matter, so your formulation is probably too wide).
As a lifelong member of the British computer society, and also an accountant, may I suggest at all computer systems should be designed with an audit trail that may be followed. I cannot imagine why this safeguard was ever discontinued. John Wyborn.
Somebody passed a law saying that a computer generated solution was to be presumed correct unless other evidence is produced quarrying it. That’s why we got the Sub-postmaster Scandal. One hopes all this will be resolved before the development of quantum computing
Your opening paragraphs are an accurate summary of the position so far as they apply to the criminal law. The position in civil proceedings is more complicated. In 1992 my firm acted for the successful claimant in R v Coventry Justices, ex parte Bullard, a case concerning the admissibility of computer records said to show that the claimant had not paid the community charge (the poll tax) imposed by the local authority. At the time the Civil Evidence Act 1968, which contained provisions somewhat similar to those in PACE, did not apply to magistrates’ courts. The outcome was that the regulations relating to the evidence required to establish non-payment of the community charge were amended to allow local authorities to rely on computer records. Although the proceedings were civil in nature, detention in custody was the ultimate sanction for non-payment.
In simple terms:
• When a computer’s hardware goes wrong, it is highly unlikely to change the orderly execution of a computer’s instruction set that represents the software running. Typically, gross symptoms occur, for example the machine generating a lot of garbage symbols in its output, which will be patently obvious,
• When a computer’s software goes wrong in the sense that an unexpected outcome is delivered or the program aborts or “hangs”, for example, apart from the latter, there is no obvious way to know if the internal software logic has been wrongly programmed or some software procedure used to correct data or other errors has inadvertently created further errors.
Having to prove that computer hardware is functioning correctly is a ridiculous requirement.
Having to prove that software is working perfectly (i.e. exactly as specified and 100% bug free) is, in practice, impossible expect for the very simplest and often trivial programs. There used to be an 1970s endeavour called program proving, which would theoretically prove that commercial software was bug free. This was abandoned. It can work in very limited circumstances to show, for example, that a self-contained algorithm does function reliably and predictably. Program testing is the way that software is checked for bugs. This is normally done nowadays where test data is created to “stretch” the program and then the output is verified to check that it corresponds to the specified requirements. This cannot and does not eliminate all bugs.
To summarise simply: The probability of a computer hardware fault causing software instructions to malfunction is near zero. The probability of software containing bugs is near to 100%.
Hope this quick skim across the top helps.
Many thanks indeed.
Such erudition from all! Genuinely, I say “thank you”, gentlemen. I am on this circuit to LEARN
On the “footnote on recusal”: it is important to note a distinction which is not clearly spelt out in Conor Casey’s Policy Exchange paper to which you link but, reading between the lines, he clearly accepts - and is plainly right.
That distinction is between a general question (“has the AG advised on issues where he has a conflict”) and a specific question (“has the AG advised/been asked to advise on issue X, where there seems to be a conflict”).
The problem Casey skates round is that at least one of the questions that he cites is a specific question (Jenrick’s “has the AG recused himself from advising on any matter involving Gerry Adams?”).
If the AG answered that positively (“yes I have recused myself”) he would in effect be disclosing that the Law Officers have been asked for advice on the Adams matter.
That would breach the convention (rightly accepted by Casey, for the reasons he gives) that the Law Officers do not say whether they have advised on any particular matter.
Yes, that's an important point. Thank you.
Could he reply: "I can't say whether or not I have recused myself in any specific case but if I am (or have been) asked to advise on Gerry Adams then I would recuse myself"?
Yes I suppose that that’s right (although I am not sure as a matter of fact whether he would be precluded from ever advising on anything to do with Adams just because he acted for him on one matter, so your formulation is probably too wide).
Yes, I agree. Thanks again.
As a lifelong member of the British computer society, and also an accountant, may I suggest at all computer systems should be designed with an audit trail that may be followed. I cannot imagine why this safeguard was ever discontinued. John Wyborn.
Seems a good idea. Was it cheaper not to bother?
Somebody passed a law saying that a computer generated solution was to be presumed correct unless other evidence is produced quarrying it. That’s why we got the Sub-postmaster Scandal. One hopes all this will be resolved before the development of quantum computing
Your opening paragraphs are an accurate summary of the position so far as they apply to the criminal law. The position in civil proceedings is more complicated. In 1992 my firm acted for the successful claimant in R v Coventry Justices, ex parte Bullard, a case concerning the admissibility of computer records said to show that the claimant had not paid the community charge (the poll tax) imposed by the local authority. At the time the Civil Evidence Act 1968, which contained provisions somewhat similar to those in PACE, did not apply to magistrates’ courts. The outcome was that the regulations relating to the evidence required to establish non-payment of the community charge were amended to allow local authorities to rely on computer records. Although the proceedings were civil in nature, detention in custody was the ultimate sanction for non-payment.