Wednesday, June 06, 2007

A Bit's Worth

There are 10 types of people in this world: those who understand binary and those who don't. ~Author Unknown

Someone at Discover magazine, evidently with a deadline to fill and no real topic at hand, has calculated the weight of the data on the Internet on an average day. I wouldn't dream of spoiling the article for you by telling you the answer ... well, yes I would, because it's one of the slightly silly and somewhat meaningless exercises undertaken by popular science media.

It's 0.2 millionths of an ounce. And most of it is pornography.

Okay, I made the bit about porn up, but the minuscule number is legit. It's so small because the "data" is actually electrons. And electrons, to use a technical term, are teensy.

Technical folks (and scientists, for that matter) are fond of coming up with such things, partly to show how clever they are and partly to "put things into an understandable perspective." In this case, it's an attempt to give some idea of the masses of data floating around the internetworks of the world. Of course, it does the reverse by coming up with a weight that is less than that of a grain of sand.

To tell the truth, I'm not sure what it proves. The process of determining what the weight is, on the other hand, is somewhat interesting, so perhaps that was the author's point.

It brings to mind another computing conundrum: How long is a bit?

Back in 1995, I was lucky to be learning my trade by working with a number of extremely talented geeks whose depth of networking knowledge constantly amazed me. One of them, who I will call Ed (because that was his name; I have no imagination) wanted to get out of the contracting game we were all in and get a job with a real networking outfit. When Network General (now part of McAfee, I believe) invited him to interview, he was in heaven. Since most of what I know about networking came from Ed, I was pretty confident that he would get the job.

He came back from the interview looking like his best friend had just died. It turned out to be fiendishly difficult, because the kind of jobs that Network General was filling require a level of knowledge that goes to the fundamentals of how bits are processed. Despite Ed's demonstrated expertise, he realized that there was still a lot he had to learn. One question they asked, though, really chapped him.

What, the interviewer asked, is the physical length of a bit?

Now, a bit, as in the case of the weight problem above, is basically an electrical signal, so it's made up of electrons, right? So a bit is, as stated above, teensy, right? According to the interviewer, it wasn't. But, what was really aggravating is that when Ed asked what the answer was, the interviewer said, "It depends."

A few months later, I was working another gig and was introduced to a another member of the company who was supposedly a networking genius, so I asked him about the length of a bit. He looked at me sagely, and said, "It depends."

"Depends on what?" I cried.

"Oh, the bandwidth, the cabling, that sort of thing," he said. So I came back at him with specifics: 10 megabits per second (mb/sec) over copper wire. He hemmed and hawed about how there was more to it than that and wandered off.

Well, campers, it turns out that there isn't more to it than that, so apparently he didn't know the answer either. The reason I know what there is to it is that I found the answer.

David James Clarke wrote books about Novell networking for years. To break up the monotony of terminology and installation methods, he would inject little puzzles and odd facts. One of those odd facts was the method for calculating the length of a bit. Here's how it's done.

A bit is made up of electrons which travel at the speed of light, which is 300,000,000 meters/sec (using meters makes the numbers easier). When electrons travel through a medium, like copper wire, they are slowed down or attenuated. Let's say we're using copper; the attenuation rate is about 80%, which means we multiply our speed by .8, yielding, 240,000,000 m/sec. Now a typical modern network has a bandwidth of 100 mb/sec or 100,000,000 bits per sec. So, if we divide 240,000,000 by 100,000,000, we get 2.4 meters per bit.

That's right. A bit is about 7.8 feet long. This is supposed to be a real phenomenon. I'm not so sure about that. For example, that same bit is 24 meters long at 10 mb/sec and 0.24 meters at 1000 mb/sec. Now, if you think of clay being pushed through different diameter dies, it will be thicker and shorter at larger diameters. But electrons aren't clay. Also, moving at 80 % of the speed of light, there would be length contraction as well, so what is the real length?

Frankly, I think it's not a physical phenomenon because a bit is not a physical entity but a stream of electrons. Now the stream may be some physical length, but I've never found anyone who would relate the bit length to that.

However, the next time that guy from IT looks at you like you're an idiot because of something you've done to your computer, challenge him with the bit length question. When he says he doesn't know and asks what the answer is, just lean back in your chair, look smug, and say with great wisdom:

"It depends."

No comments: