Right-wingers Don’t Understand Software
Well, at least one right-winger doesn’t. Here’s a howler from Taylor Dinerman at Pajamas Media explaining why we shouldn’t bother to test weapons systems before buying and deploying them:
Certainly testing sounds reasonable. Why not make sure the stuff works before blowing billions on it? But the testing fixation ignores that, like software, most successful weapons systems are best debugged after being deployed. And some weapons systems were never tested at all before deployment.
(Emphasis added)
Nobody, but nobody, in the software world believes you should test or debug after deployment. Testing as early as possible is the mantra of modern software development. Test first programming is the most significant improvement in software development in decades.
Debugging after deployment is a clear indication of failure in the software development process. It happens, but smart programmers do everything they can to make sure they have to do as little of this as possible.
Even in the classic waterfall model, testing and debugging happens before deployment. That didn’t work very well so we’ve now moved testing even earlier in the software development life cycle, and it works much better.
Is he seriously suggesting a “Ready, Fire, Aim” approach to weapons development? At least software doesn’t kill people when it’s buggy. (Well, at least not most of the time.) A defective weapons system needs more and earlier testing than software, not later and less.
I’ve already commented on his blog, but perhaps some of the right-wing developers out there would also care to set this lunatic straight.
December 3rd, 2006 at 1:45 AM
Left-wing Bloggers Use Sweeping Generalizations
Well, at least one does.
December 5th, 2006 at 12:09 PM
Left-wingBloggers Use Sweeping Generalizations.Well, at least all of the bloggers I have read.
December 5th, 2006 at 2:12 PM
well I think it’s safe to say he put his point rather badly from an engineer’s point of view because he’s a layman writing for laymen, not an engineer. To leap from that to “Right Wingers Don’t Understand Software” is, well, quite a leap.
I think the point he is making is a valid one, and I think it applies to you, as well as to the members of Congress the article addresses.
The point is, as he says: “In short, Sen. Levin and other longtime opponents of missile defense plan to use “testing†– set to an unrealistically high level – to stop missile defense.” Similarly, you are focusing on one sentence and making fun of it rather than responding to the article seriously.
If you read the entire article in this context, it makes sense. The examples he gives are true and valid examples.
Rather than come right out and oppose missile defense systems on strategic grounds, which they know won’t play well with the majority of voters, they kill it with the death of a thousand cuts. They believe this deceptive approach is quite reasonable because they know better than the voters what’s good for them. This arrogance is quite common. Your arrogance in proclaiming, essentially, “Right Wingers are Stupid” is quite similiar.
December 7th, 2006 at 11:07 AM
Sounds about right. I don’t buy any of the excuses posters are making for this idiotic comment. (“Debugged”, to me, is a bit too little for a system that’s never actually worked in any of the non-jiggered tests so far. Something that’s failed even the most basic proof-of-concept tests doesn’t require debugging; it requires fundamental changes.)
That sad part is I hear very similar comments coming from IT project managers all too frequently — so, yes, there *are* people in the software world who seem to believe that, mind-boggling as it is.
December 22nd, 2006 at 7:18 AM
well, I’m certainly a fan of functional tests in a staging system. Maybe the author implied “The best way to test weapons systems is in small wars”. Gulf war I showed up bugs in Patriot; The falklands war exposed the UK navy to exocet related defects.
In the near future, a ballistic missile engagement with north korea would test out the ABM shield without exposing more than seattle to destruction. and you could argue that Iraq/Gulf-war-II is actually Iran beta testing its algorithms to deal with the forthcoming Gulf War III.
The issue w/ in-field testing of weaponry is that nowadays the design cycle of products is too slow for feedback to be incorporated into the product before that specific conflict is over, and that the failure modes of the system come in too high up the “loughran scale” of system engineering disasters. A Level 8 error “loss of small country”, or even L7 “loss of city” is much more significant than an L1 event “loss of unsaved work”.
-steve
July 7th, 2008 at 11:47 AM
Hmm. I wonder if this was a simple typo (the link has rotted)? For “is best debugged” read “is at best debugged”, a very different matter.