Productivity Gains With KDE4

Honesty’s a good thing, usually. In particular, it’s usually a good thing in software engineering, in which most of what we do is digital and repeatable, is either one thing or the other; this fosters a culture of honesty in the same way that it does (or should do) in science, as Feynman points out in Cargo Cult Science. Unlike in a courtroom, or even in a courtroom drama, software engineering rarely comes down to one person’s word against another. (Well, unless your co-workers are sociopaths.)

But there are situations where more honesty isn’t a Good Thing. One example, in fact, is a courtroom drama: if you’re writing such a thing, the principle of pure honesty would have you title it something like “Not Guilty of Murder” — whereas, in fact, letting people know the verdict before they’ve seen the piece robs it of its whole point.

Now KDE3 came with a desktop toy called KPat, a patience game which (among others) includes an implementation of the same “Klondike” patience found in Windows Solitaire. The KDE3 (3.5.10) version of KPat had a feature where, if it detected you’d got into a situation where it was impossible to complete the hand, it would stop the game and tell you so. Naturally, such a feature has to be extremely conservative: it must stop the game only when it can be proved that forward progress is impossible. And in fact the algorithm in KDE3 KPat was very conservative indeed: it kicked-in so rarely that seeing the message always came as a surprise, even on games you already knew you’d lost.

Occasionally I’d idly wonder whether the lost-game detection could be improved — but then I realised that actually, if you improved it enough, you’d eventually get to a situation where the game detects, and tells you, the moment you’ve made a move that leads only to dead-ends. “OK, you’re an idiot, bye, next.” There’d be no point playing the game at all.

And yet, in the KDE4 (4.3.2) version of KPat, that’s exactly what’s been implemented. In the status bar, the whole time, is one of two messages — either “Solver: This game is winnable” or “Solver: This game is not winnable in its current state”. Any time you make a move that makes the first message change to the second, you soon reconsider!

And so, alongside the huge amount of work on the graphics (the original neat bitmaps have become huge and flouncy SVG images; the codebase diff is huge even ignoring the Solver; the whole thing has unexpectedly acquired an Ancient Egyptian feel) the developers have completely ruined the actual game. All the time you’re playing, it’s as if a stern examiner is watching over your shoulder, always ready to lean forwards and intone “Now I don’t believe you wanted to do that”. Worse, unlike the KDE3 algorithm, which just summarised information you could already see, the examiner can see the cards that you can’t, making it an eerily omniscient guide, not to mention a shocking cheat.

Solving patience is a great technical achievement. And it’s certainly scrupulously honest to tell the player exactly what the prospects of success are. But it’s a technical achievement that shouldn’t have been achieved (or shouldn’t be present in the game itself, even though it can be turned off), and a case where honesty is definitely not the best policy. And one of the best little time-wasters in KDE has, effectively, been eliminated.