Campus folklore suggested many methods for beating the phone companies, but most didn't really work very well. One of them was to make the weekly call home on a different day and from a different campus phone every time, the idea being that this would make it harder for the phone companies to track you, so you'd only get generic student ads for fast food places and the like instead of ads targeted at your specific weaknesses. Tim found that this method wasn't worth the trouble, but he did use the one that said to avoid local languages on long-distance.

"Why do you always speak English with me now?" his dad asked him in Dutch. He was charmingly chauvinistic about his tiny corner of the European Union, even though he worked in Amsterdam, where hardly anyone spoke Dutch anymore. (His name was Joost. His son's name used to be Kuifje.) "It saves money," Tim replied in English. He figured this would impress his dad, who was not only a former businessman but was also paying for his studies in Toronto, including the phone bills.

"That doesn't make any sense," Joost said in Dutch.

"I didn't believe it at first either," Tim said in English. "But I tested it and it works. Try it yourself. Local language costs almost twenty percent more."

That boy's a real scientist, Joost thought to himself (in Dutch).

But Joost knew he didn't need to try it himself. He didn't need to be a scientist, begging for scraps of truth from God. He was Assistant Undersecretary of Commerce in the European Union Central Government. He had God's phone number.

"Of course your son is right; I can't believe that you really didn't know this," Trubetzkoy said in extremely rapid English. Trubetzkoy controlled fifty-three percent of the company that controlled seventy-six percent of the companies that controlled ninety-eight percent of the global telenet industry.

"But --" said Joost in English. Then he heard the advertising warning bell and stopped talking, as he had been conditioned. It was the one from that nanotech lobbyist.

Yet almost as soon as it had started, the woman's soft voice was cut off in mid-spiel. Joost didn't have to ask who had done it.

"You in Central Government, and still not ad-free?" Trubetzkoy rapidly clucked his tongue. "I'm shocked; you really have no excuse not to understand the system; you must learn to understand the system or the system will destroy you."

"Sorry," said Joost in English.

"OK, I've got to go; just read this." Paper started streaming out of Joost's machine as Trubetzkoy went off-line.

Joost expected a top-secret document, but it wasn't. It was a reprint of an article that had first appeared over a year ago, in English, in a telenet trade magazine. Who could keep up?

The article explained why Tim's phone bill method worked. The key was that the telenet companies liked money. True, after switching to an all-data-packet system without passing the savings on to users, profits had soared. And yes, phone advertising had been an enormous success, surprising even the most cynical in the industry with what the public would put up with. But now there was a new stock-boosting revolution on the horizon, one that was made possible, as usual, by new technology: language compression.

The roots of the idea went back to the death of dead air. People talking on the phone aren't always talking. Why pay to transmit silence?

Shunt resources over to another line during the lulls and there's another phone call you can sell.

And don't neglect the tiny lulls in the middle of words. Sounds like "t" and "p" are mostly silence; callers never notice when part of their "nighttime" is used to send a phoneme or two of somebody else's "love."

Even speech can be turned into silence and thus into money.

Speech is sound and sound is waves and waves are often alike. Clip out the duplicate waves at one end, send the clipped bits in code, put it all back together at the other end, and a half-hour gabfest becomes a ten-minute chat, with the leftover time sold to the next user.

"I know this already," Joost said to himself in English as he turned past an ad to the next page. And then he read some more.

Recently, however (the article went on), the industry had realized that purely acoustic compression had missed the most important source of redundancy in phone conversations: conversation itself. It wasn't much trouble snipping out the hmm's, uh-huh's and you know's, but to achieve maximum efficiency, the telenet companies decided to rebuild human language from the inside.

When human language was a cutting-edge technology in the Neolithic, the fact that it was packed with redundancy was a great design feature, since the Neolithic was a noisy place. In this age of fiber optics and data packets and complex systems theory, however, redundancy is mostly just a waste of money.

Any language will do to illustrate this. Take an English sentence for example, a perfectly ordinary one, nothing fancy: "Yesterday I went to the store, and today I am going back". Sixteen syllables of deathless prose. What could be more eloquent? This could: "Yest I go store, tod back". Six syllables, and no loss of information. "Yest" is enough to indicate "yesterday" (the less common "yesteryear" could be "yesty"), and that's enough to indicate the past tense, and since people usually go "to"

the store, the "to" is redundant (only if they go "in" or "by" it would you need a preposition), and what's the point of the "the" anyway? And this is just what you find with an unusually plain sentence.

Statistically, the typical English sentence consists of even more useless fat, fat that could become profit if it were trimmed away.

The problem was that there was no single language trimming algorithm that would work for all languages, since each languages packs its cellulite in different places. Russian may have no "the", but it has three genders and six cases; Chinese may not have any genders or cases (nor tenses and suffixes for that matter), but it's filled with compound words where each part just repeats the meaning of the other; and so on for every other language in the world, an efficiency consultant's nightmare.

It was decided, therefore, that the most economical solution would be to compress just the most popular languages, and the obvious place to start was with the most popular language in the history of the world.

Temp linguists were hired to create a maximally efficient English-like language, dubbed English-Minus or E- for short, and temp engineers were hired to create software to compress regular English into E- and decompress it back out again. Then the linguists and engineers figured out how to translate E- into and out of a universal semantic code that could be sent essentially instantaneously. Voice quality information was encoded separately, so when your sister's lisp or your psychiatrist's accent came over the line, you had no idea that the whole thing was synthesized.

Once it was up and running, the language compression system saved the telenet companies such a bundle that this time they felt generous enough to pass some of the savings on to users. Of course, this was only if they spoke English (or some big local language like Chinese or Spanish which were also worth the cost of hiring linguists and engineers). And here was a tip that Tim had missed: users also saved if they stuck to common topics that were readily translated into the universal semantic code. This didn't violate anybody's freedom of speech; you could still discuss things as far off the beaten path as you wanted. It was just that you'd have to pay extra for it.

"No wonder Tim has sounded so strange lately," Joost said to himself in Dutch, defiantly in Dutch. "He's being computer-generated."

"Isn't that what I already told you?" Tim said on the phone a moment later. He, or whatever it was, still spoke English. "I'm saving money, so why does it bother you?"

Joost hesitated, and wondered who got his dead air.

"I don't know."

He stopped again before he knew why, as he had been conditioned.

The ad bell. This time it was one he had never heard before, but it seemed familiar. It was the ad he'd seen in the article: Dutch speakers were wanted for a temp project in Toronto.

When it was over he asked if Tim had got the same ad.

"Yeah, and it's weird. I already work there part-time."

The machine in Joost's office spit out another piece of paper. It began "We regret to inform you...." The rest was redundant.

Father and son were reunited, live and in person, the following week. Joost rented an apartment outside Toronto an hour by subway from the Local Language Compression complex and he spent this hour trying to believe Tim's words of encouragement.

"My generation doesn't speak much local language anymore, so they really need old guys like you. Your job should be pretty secure. At least until the technology advances again."

His new office was a lot smaller than his old one, but it had a machine with buttons on it in Dutch. It also had explanatory screens that scrolled open when needed, and they explained in Dutch. His new supervisor came in from behind and talked to him. He talked in Dutch, more precisely in the musical tones of Flemish, the Dutch of the Belgian Region. Joost turned from the machine with a smile but the supervisor had gone to greet another new arrival. They needed a lot of Dutch speakers working in parallel on this.

The job was to hand-compress Dutch call packets into Dutch-Minus.

Apparently for a dying local language it was cheaper to have semi-skilled labor hand-compress on-line than to hire more linguists and engineers to automate. For five hours a day he listened to snippets of conversations between callers who seemed about his age, snippets just long enough for him to catch the basic meanings of sentences but not the whole conversation, and as he listened he pressed buttons labeled with guesses about common words and phrases and grammatical properties like tense that he expected to come later in the sentence. He wasn't as fast as a computer, but each button press meant that the acoustic compressors could be more savage with the rest of the sentence, and each was added to a growing database that would make possible future automation that much easier.

He did this for five hours a day, as quickly and accurately as he could, because he hoped this was all just a test and he would get his old job back when he had proved that he understood the system, but also because he felt sympathy for the hundreds upon hundreds of anonymous voices which passed through his earphones every hour, discussing shoe prices and weddings and pains in sides and people named Mirjam and the secret of a good pie and the phone bill and strange dreams and a small multicolored insect someone's daughter had brought home from the park, and he wanted to do a good job so they could express themselves in the language they chose to speak because it was the language and the speech of their own voice and mouth and mind and heart and living breath.