Monday 14 January 2013

Systems ≈ Random?


I don’t quite believe that yet, but I’m getting closer.

Blogger has a ‘known problem’ at the moment which makes it difficult to insert a picture in your post.  I won’t go into details, because they’re boring and nerdy (and they’ve identified a way round it,  which seems to work, whilst they fix it) – the thing is that it only happens if you use a particular browser.   That’s like saying that an identical can of beans will have different contents depending on how you carry it home from the supermarket.

The point that worried me was the one jokily expressed in this post’s title: that the behaviour of some computer systems increasingly seems to approximate to randomness.  Like I said, I still just about refuse to believe this, even though Blogger regularly shakes my faith.

Anyway, in support of my probably misplaced optimism, I offer you the following true story.

Not that many decades ago, I was involved in developing and testing a new computer application, which duly went live.  It worked fine for weeks, until suddenly a particular transaction caused it to crash catastrophically, dragging several other things down in its wake. 

We pored over the content of the guilty item and could see absolutely nothing unusual about it, and so handed it over to the lead programmer, Anjam.  Two hours later, he came back with the answer.

“This only happens,” he explained, “when the length in bytes of the resulting record is an exact multiple of 256, plus one.”

Over his shoulder, as he walked away, he grinned.  “Sloppy testing.”

15 comments:

  1. Tell me about it. I have spent weeks "de-bugging" other people's software systems and I know absolutely zilch about programming or what makes computers tick. But I can tell them when they go wrong or are illogical and try to explain the randomness of their going wrong. This is apart from the ineluctable inanity of some of the stuff they try to make us "lesser" mortals use. I suspect most of it is also down to sloppy testing.

    ReplyDelete
  2. I used to have a slightly superstitious approach to programme behaviour, ie, 'last time this programme worked I had just turned off the printer - perhaps it doesn't like the printer'.

    When you think of the numbers of programmes that are written it seems possible that there is a random element attached to the errors the the programmers make? (I'm imagining a kind of electronic butterfly flitting through the ether and occasionally pollinating one programme with the error from another - Oh too much Pratchett in my impressionable middle age)

    ReplyDelete
  3. I usually assume it's something I've done wrong.

    ReplyDelete
  4. I quite like this butterfly idea.

    ReplyDelete
  5. It's just twiddling init? Blogger's twiddled with something that was working just fine. And they know they've twiddled. They just can't admit the twiddling. In fact they're probably twiddling overtime by now.

    And you know, ahem, other browsers are available

    ReplyDelete
  6. Multiples of 256 plus one. Ah. I thought that would be it.

    ReplyDelete
  7. I usually know it's something I've done wrong...

    ReplyDelete
  8. Reminds me of Millennium Bug Paranoia. That didn't cause many deaths.

    Why are the Large family so in enamoured with the word "ineluctable"? Just wondered....

    ReplyDelete
  9. Regardless of the history of any given bug, let's get this straight: IT'S THEIR FAULT. Not ours. We didn't ask for it. Exactly, Macy - twiddling. I think unsolicited 'improvements' should come with an extended warranty, with penalty clauses.

    The butterfly doesn't hold water. (Eh? Cancel that image!) That would be a virus. Those ouanqueres aren't that clever.

    Macy - your last point is true, but how do I know their next twiddle won't hit Firefox, Chrome, etc? (and please, nobody use the A word...)

    Rog - once I start something in the L family, it's ineluctable.
    (Oh, and the millennium thing wasn't a bug, it was a planned design feature. Just, nobody thought those systems would last that long.)

    ReplyDelete
  10. Sorry, that should of course have been 'ouanqueures'.

    ReplyDelete
  11. Tim, youare such a ouanderfully ouicked ouag.

    ReplyDelete
  12. Hola! I'm back.
    On Chrome.
    Because Firefox is now having problems reading the emails on icloud......
    but you see it can be done!

    ReplyDelete
  13. All this is a bit beyond me but I know when proof reading/testing a friend's new website last year, I was told by the web designer that no-one uses IE any more so he wasn't going to make the website compatible ..... excuse me, I told him, anyone over 60 learned the internet on IE and probably aren't even aware that there are dozens of other browsers! He sulked, but did it!

    ReplyDelete
  14. Well, that web designer needs to get past Key Stage 1. And it makes my point - the browser is just the conduit. If they need to make their apps 'compatible', there's something seriously wrong somewhere. I've never used anything but IE, and see no reason to.

    ReplyDelete
  15. Having done some more research and uncovered the existence of something called 'mshtml' (decode the first two letters for yourself), yes, there's something seriously wrong. Oldies will remember the concept of 'Unix', and what became of that.

    ReplyDelete