Do Aweb and Fujitsu share a common 'feature'?

If you've ever been slightly annoyed by the duplicate/triplicate comments sometimes posted here ...

Didn't find your answer?

Before the moderators, or even lawyers, jump into action, I'm not accusing Aweb of any malfeasance ... but digging into reports of what may have caused the P.O. errors within Horizon brought to light a similarity with a well-known defect in the software used by Sift.  You know, the one where nothing seems to happen when you hit the 'Post reply' button ... so you try it again (and again), until it wakes up and unfortunately now posts N copies of your comment.

As I understand it, this is because 'posting' is a 2-stage process:

1st your words are sent to a 'holding bay' (a bit like limbo) where it is inspected for egregious behaviour or wording that contradicts one of the many rules of the site.  Initially this 'vetting' is done by software, but it can be passed to a 'detention bay' and held there for later consideration by a human; all of which is fine & dandy, and imperceptible for 99%+ of submissions because the software is fast and, if happy, your words proceed to the final stage.

2nd a suite of automated processes formats your words (including noting to whom it is a reply and date/time stamping it etc), and 'publishes' your comment.

Why is all this relevant?  Because IT systems are rarely 100% reliable (even at NASA), and it is not unknown for a hiccough to occur during the 1st stage - at which point (and I'm guessing here but based on a common failing of many systems) it makes a copy of the 'problem' file and tries to process that copy in case the issue was data corruption in the original file.  So there are (in this case) now 2 identical files - since there wasn't actually any corruption - which will eventually be processed with identical results ... and, if successful, the original set of words will therefore be published twice.


Moving on to Horizon - where, according to Mr Bates (ably assisted by IT supporters), whenever the Horizon system encountered a transaction that (for whatever reason) it failed to 'complete' (commit to the database in nerd speak) ... the same kind of problems emerged.  To use more of the jargon, the failed transaction wasn't 'rolled back' but simply repeated ad infinitum ... until it achieved 'success'.

Unfortunately (I'm in polite mode here) that meant that N copies of that transaction would eventually be committed to the database (see the Aweb example to avoid repeating myself) ... including of course the N-1 transactions that weren't actually real.  But unlike Aweb, the consequences were catastrophic for the people assessed as liable for any (non-existant but reported) shortfall shown by the system!


All of this was exacerbated, indeed made possible, by the fact that Horizon (despite the claims made by the Post Office) was never designed as or even requested to be a true Accounting system.  In essence it was no more than a glorified electronic till, dressed up as the accounting arbiter but not even incorporating proper itemisation of transactions let alone anything approaching double-entry bookkeeping.  So ... not designed to do what was needed and therefore incapable of highlighting what should have been the most obvious errors of duplication - but no-one in charge understood this (or worse, just didn't care)?

Apologies to Aweb for indirectly sullying their name, but I thought the example might make it easier to understand what has been going on at the P.O./Horizon ... and maybe why those in charge are still so reluctant to admit the details, which would make culpability that much harder to avoid.

Replies (10)

Please login or register to join the discussion.

By Open all hours
11th Feb 2024 16:08

1) Absolutely right to at least raise the matter. I have long been frustrated/puzzled by how easy it is to post duplicates on here.

2) A question I have not been able to find an answer to - When Mr Osborne announced the end of the tax return was he reading a Fujitsu script?

3) Postmasters look out, it’s that week…

Roses are red
Violets are blue
It’s not me, IT’S YOU,
love, Fujitsu

Thanks (0)
By blox
11th Feb 2024 22:38

Before holding bay it actually has to hit the server where the back end of Aweb is running. And each “Post reply” click may send the whole data yet again. It could even be that your second Post reply hits the server before the first.

This is not just the issue with Aweb. It happens with many websites and this is why some payment processing apps give a message to click on Pay button only once or that you could be double charged.

Thanks (0)
Replying to blox:
By FactChecker
11th Feb 2024 23:04

True ... although I did say "I'm guessing here" before suggesting that sometimes "it makes a copy of the 'problem' file and tries to process that copy".
And you're quite right that many such systems somehow, from time to time, allow a subsequent submission to hit the server before the first attempt - which is why RTI has been known to get 'out of sync' when it has polling issues and the sender decides to take advantage by making a correction (which then gets overwritten without the sender's knowledge by the late arriving first submission)!

But my main point (if it's got lost) was not the 'issue' with Aweb - it was simply using the Aweb example as a point of reference that might aid understanding of the similar fault within Horizon, which hasn't yet received much publicity but obviously had catastrophic results.

What I want to know (mad optimist that I am) is who in Fujitsu and who in the P.O. knew this and kept quiet?
It was far more important even than the lie that transactions couldn't be made or amended remotely by Fujitsu staff, as it undermines the central tenet of reliability - without which the basis for most prosecutions would've evaporated.

Thanks (2)
David Winch
By David Winch
12th Feb 2024 10:10

OK, here's a moral dilemma (maybe).
You are a postmaster/postmistress. You have to 'balance' on the Post Office system (after close of business) today or the system will not permit you to open for business tomorrow. The system is showing a cash shortfall figure which you feel cannot be correct. You are obliged to make good that shortfall out of your own pocket. What do you do?
If you decide to put in a 'fake' transaction to 'balance' the system tonight which you can reverse out tomorrow, which will give you a week's breathing space to sort out the issue - is that a reasonable thing to do? Or is that a dishonest transaction entry making you guilty of the criminal offence of 'false accounting'?

Thanks (2)
Replying to davidwinch:
By FactChecker
12th Feb 2024 16:32

Not sure it's strictly a moral dilemma ... although highly pertinent, as I'm sure you know, in terms of handling the 'shortfall' on a day-to-day basis (rather than the cause and associated culpability which is where I was focussed).
I guess I'm just being pedantic in that it was more of a catch-22 (or colloquially the 'heads I win, tails you lose' offer by management to sub postmasters/mistresses).

Alan Bates was in a tiny minority (but not alone) in refusing to accept either 'offer' - and indeed refusing to sign anything that could be seen either as confirmation of the shortfall's existence or of his responsibility for whatever was going on.
This of course meant that he was locked out of the system and eventually also out of his own premises, but it put the onus on the P.O. to take any action regarding fraud or theft - which they unsurprisingly declined to do!

That's what you might term the nuclear option, which most wouldn't or couldn't take - some because they became convinced they must have 'done something wrong', and many because they couldn't take the financial or reputational hit (or both).

But it's interesting because it sort of takes us back to my original point.
If the P.O. were able to 'accept' a lack of 'shortfall repayment' without proceeding to the courts, then that's a smoking gun indicator that they knew the 'missing money' wasn't (always) real.
Which in turn not only raises the central issue of who knew (and why they let the minions get on with all the persecutions), but what they did with all this 'extra' money flowing in that didn't have a liability with which to cross-reference?

The stench is not diminishing - and, as per my main post, was pretty much baked-in before the Horizon terminals started being installed. That's the disgrace!

Thanks (2)
By D V Fields
12th Feb 2024 20:11

Have you been told you are the only one having these problems?

Thanks (1)
Replying to D V Fields:
By FactChecker
12th Feb 2024 20:24

Subtle as a Flying Mallet :=)

(which was a great album that transcended all the then current fashions - a lesson still for those in big business/government today?)

Thanks (2)
By Yeadonian
13th Feb 2024 08:09


Thanks (0)
By Tom+Cross
13th Feb 2024 11:50

Wonder if the name Fujitsu will survive much longer, after all of the water that's gone under the bridge.

I'd imagine that retribution will be; extensive and the resultant outcomes could well be permanently damaging. Let's be honest, before the ITV series, we all imagined that the software would be proven and secure.

After all of this, who would you trust, in the future? Certainly not our barons of industry.

Thanks (1)
Replying to Tom+Cross:
By FactChecker
13th Feb 2024 23:27

Maybe a re-branding is indeed called for ... they could acknowledge P Eye's part in the investigative journalism by indirectly referencing the infamous law firm of Sue, Grabbit and Runne.

So goodbye Fujitsu ... hello Fudge IT Sue?

Thanks (2)