All ye non-geeks dare not enter. Ye hath been warn'd, verily.

( Complex validation done right... )

[g+]

Doesn't it seem like there's been a lot of hacking and password cracking in the news lately, like the Sony network and other places? Last year with China's attack on major us corporations? Well, good news everyone! You don't need to worry about strong passwords, you're probably fine just the way things are!

( Admit it: you read the last part of that last paragraph in Prof. Farnworth's voice. )

[read this on g+]

I've been posting Google+ tips on my account over at g+ for a bit, and I realized just now that these are mostly for people already savvy to the basics. I know this b/c ppl I've recently invited have said, Hey, what are the basics? Ok, comin' at ya.

( Your home screen... )

There's a lot more to g+, but this is the 5-minute intro-primer beginner getting-started stuff you need to know. Fun! Win! Win-fun!

With the release of Google+, I've been thinking about privacy and social.

Before I say anything else, just to be totally clear: I don't have any special knowledge about, access to info, or insight into Google+ and how decisions were made beyond what's publicly available to everyone. I don't work on social or anything close to it. These are my thoughts and mine alone, I don't claim to be an authority, in fact I'm quite sure that those who did work on this have thought far longer and harder on this matter than I have.

( Privacy and social have a natural tension... )

I bought a book so I'd have something to read on a plane ride called Do You Think You're Clever? by John Farndon. Also, I bought it because I think I'm clever.

( Food for thought... )

You buy a particular brand of jelly from your local grocer from time to time. On the inside of the jelly jar lid is printed a number, which you come to know from your grocer is the number of that jar within the batch, which is always between 1 and the batch size.

You ask your grocer how many jars are in a batch, but he doesn't know. He does tell you that the every batch is always the same size, whatever it is, and that he places orders whenever he runs low on stock, which is fairly erratic. You take this to mean that the store receives a random sampling of jars (i.e., if the batches are large, the store is not receiving jars from the same part of the batch by coincidence of schedule).

It so happens that you have always saved the jars & numbered lids. You go home and go through the lids and find the biggest number (we'll call it N_max), so you know that batch sizes are at least that big. You would like to figure out approximately how many jars to a batch, to a confidence level of 2σ and a confidence interval of N_max/10. How many lids would you have needed to collect to get this estimate of a batch size?

Two aliens are chatting.

Humans are weird, a grating robotic noise, spewing green smoke and dribbling a caustic glue-like substance.

The other one regards the former disapprovingly. Why's that? It generates a swell of coherent light beams and airy sonic reverberations.

They like to mash things up in the wet hole in their heads. Did you know that inside that wet hole there's these bone nubs? Suddenly it expels a hard puck into the surrounding matrix at high speed. They just love grinding things up with them.

They do that because they have to. Their system has to break down food so their different meats can extract energy from it. That's the first step.

Oh. But then...what's a gumball for?

You have a 100-sided fair die and you are testing for the event that a 37 is rolled. You don't have direct access to observe the die, however, you must rely on a complicated technology that observes the face of the die and reports the result to you; when you run a trial, the system lights either a green LED ("rolled a 37") or a red LED ("did not roll a 37"). The equipment that observes the rolled number, however, is not perfect—on average, 1 out of every 50 trials the system fails to read the number properly and observes a random value equally distributed between 1 and 100.

Observing this behavior, I notice that if I replace the system described above (let's call that system S1) that observes the die with a new system (S2) that simply always lights the red LED, this new system will only report the incorrect result 1% of the time (when a 37 is rolled). Without observing the die at all, S2 is able to cut the error rate of S1 in half.

Question: is this a valid argument for doing away with S1 altogether?

[UPDATE] cgm caught an error in this problem setup! The way I set this problem up, the error rate is actually not 2%. You can see this easily if you work out all 4 of the contingent probabilities...

• roll a 37 (0.01) | report a 37 (0.98)
• roll a 37 (0.01) | report a not-37 (0.98)
• roll a not-37 (0.99) | report a 37 (0.02)
• roll a not-37 (0.99) | report a not-37 (0.98)

The astute observer will notice at this point, as cgm did, that the third contingent probability is incorrect. In the event a not-37 is rolled, the way I described the system above, S1 will most definitely not report the wrong answer—that a 37 was rolled—2% of the time. Rather, it will choose a random number between 1 and 100 and report that, which will be a not-37 99% of the time, and therefore it will report the incorrect result only 1% of the time.

(The very astute observer will further note that there is another incorrect contingent probability in the table above, but I leave it to you to figure out.)

So, how can I achieve the correct behavior I was after? I should have said that the switch controlling which LED to light is flaky, and lights the wrong LED 2% of the time. So let's pretend that's what I said all along, and we can put this whole thing behind us. :-)

My advice is: don't touch the lens surfaces unless you have to. Ways to accomplish this...

( Clean glass is good glass... )

The Domain Name System, or DNS, is the Internet system that translates human readable names in the URL bar of your browser into IP addresses that actually route to machines.

( Probing the web for fun and profit... )