The Victorian Internet (10 page)

With the use of one commercial code, for example, the following lengthy message—"FLOUR MARKET FOR COMMON AND FAIR BRANDS OF
WESTERN IS LOWER, WITH MODERATE DEMAND FOR HOME TRADE AND EXPORT; SALES, 8000 BUSHELS. GENESSEE AT $5.12. WHEAT, PRIME IN
FAIR DEMAND, MARKET FIRM, COMMON DESCRIPTION DULL, WITH A DOWNWARD TENDENCY; SALES, 4000 BUSHELS AT $1.10. CORN, FOREIGN NEWS
UNSETTLED THE MARKET; NO SALES OF IMPORTANCE MADE. THE ONLY SALE MADE WAS 2500 BUSHELS AT 67c"—can be reduced to: "BAD CAME
AFT KEEN DARK ACHE LAIN FAULT ADOPT," a mere nine words.

By 1875, the use of commercial codes was starting to get out of hand. Some codes involved some weird words, like "CHINESISKSLUTNINGSDON."
Six syllables, sure—but hardly easy to pronounce, and twenty-one letters long. The telegraph companies felt that too many
people were bending the rules. So in 1875, the ITUtried to clamp down on this sort of thing by imposing a fifteen-letter limit.
Inevitably, the result was a flurry of new codes that conformed to the new rules but use d bogus (though shorter) words like
"APOGUMNOSOMETHA."

In 1885, the rules were further tightened. A limit was imposed of ten letters per word for telegrams in code language, and
words had to be genuine words in German, English, Spanish, French, Italian, Dutch, Portuguese, or Latin. What's more, the
sending office could demand proof that a word was genuine. Again, new codes were immediately devised in response to the new
rules. Every move the telegraph companies made to try to reduce the use of codes was neutralized by the increasing cunning
of code compilers.

However, by this stage the drawbacks of such codes were becoming apparent to their users as well as the telegraph companies.
Each code word meant so much that a single misplaced letter (or dot or dash) in transmission could dramatically change the
meaning of a message.

One particularly graphic example occurred in June 1887, when Frank J. Primrose, a wool dealer in Philadelphia, sent William
B. Toland to Kansas to act as his agent and buy wool on his behalf. Using a widely available off-the-shelf commercial code,
the two men passed several messages back and forth as they kept each other informed of their transactions. But things went
horribly wrong when Primrose sent a message explaining that he had bought 500,000 pounds of wool. The words "1 HAVE BOUGHT"
were encoded by the word "BAY" in the commercial code, and the amount 500,000 pounds by the word "QUO," SO that "1 HAVE BOUGHT
ALL KINDS, 500,000 POUNDS" became "BAY ALL KINDS QUO."

This message was incorrectly transmitted to Toland as "BUY ALL KINDS QUO," possibly because the Morse code for "A" (dot dash)
differs by only one dot from the Morse code for "u" (dot dot dash). As a result, Toland assumed he was being instructed to
"buy all kinds, 500,000 pounds" and duly started to buy half a million pounds of wool. By the time the mistake had been uncovered,
the market had turned and Primrose ended up losing $20,000. He tried to sue Western Union, the telegraph company that had
transmitted the fateful message, but he lost because he had failed to ask for the message to be verified—an optional service
that would have cost him a few cents extra. Eventually, after a lengthy legal battle, the Supreme Court ruled that he was
entitled to a refund only on the cost of sending the original telegram, or just $1.15.

To prevent these sorts of errors, new codes were devised where words specific to a particular industry were excluded from
use as code words to avoid confusion, and all remaining words were chosen so that they all differed from each other by at
least two letters. That way, if one letter got garbled in transmission, there would be no danger that it would be taken for
another code word with a different meaning. Special look-up books were provided to help with error correction.

However, the number of genuine words of ten characters or less that differed by at least two characters from all other words
was very small, so once again the code compilers started to bend the rules and use deliberately misspelled words. This was
technically not allowed, because code words were supposed to be genuine words in one of the allowed languages, but the code
compilers knew that the telegraph clerks could hardly be expected to know how to spell every single word in all of the allowed
languages.

But by 1890, the ITU had got wind of this ruse too and decided the only solution was to compile an official vocabulary of
all permitted words; any word not in the official vocabulary would then be charged at the cipher rate. In 1894, the first
edition of the vocabulary was published; it contained 256,740 words of between five and ten letters, drawn from each of the
eight permitted languages. But it was widely criticized—not least because so many common words were omitted. So the ITU scrapped
it and decided to try again. Work started on a new vocabulary containing millions of words, but the plan was abandoned when
the impracticability of printing thousands of copies of the vast vocabulary, and getting telegraph clerks to laboriously check
every word of every message, became apparent.

In other words, as fast as the rules were changed, new codes were devised to get around them. And eventually users got what
they wanted—the ability to send coded messages.

a
PARTICULARLY important use of codes was by banks. Worries about the security of telegraphic money transfers were holding
back the development of on-line commerce ("The opportunity for fraud has been the chief obstacle," declared the
Journal
of the Telegraph
in 1872), so banks started to rely on sophisticated private codes to ensure the safe transmission of money. Although there
were existing schemes for transferring money, they were insecure and depended on a high level of trust between both parties
and telegraph operators at each end. There was clearly a need for a more secure system, which would unlock a whole new market
as people in need of money in a hurry turned to the telegraph.

In 1872, Western Union (by then the dominant telegraph company in the United States) decided to implement a new, secure scheme
to enable sums of up to $100 to be transferred between several hundred towns by telegraph. The system worked by dividing the
company's network into twenty districts, each of which had its own superintendent. A telegram from the sender's office to
the district superintendent confirmed that the money had been deposited; the superintendent would then send another telegram
to the recipient's office authorizing the payment. Both of these messages used a code based on numbered codebooks. Each telegraph
office had one of these books, with pages containing hundreds of words. But the numbers next to these words varied from office
to office; only the district superintendent had copies of each office's uniquely numbered book.

A running count was kept for each book, and each time a money transfer telegram was sent, the next word in its unique numerical
order was sent as one of the words of the message. Another page in the codebook gave code words for different amounts in dollars.
And a special password, known by the superintendent and the operators at each office, also had to be included, sometimes as
the first word in the message, and sometimes as the last word. The system was deemed to be secure enough that up to $6,000
could be transferred between fifteen designated major cities "to meet the occasional exigencies of businessmen."

The service soon became very popular, and by 1877 it was being used to transfer nearly $2-5 million annually in 38,669 separate
transactions. "This service, which meets a vast demand from parties caught in unexpected conditions of loss or embarrassment,
is one of the greatest boons of our modern civilization," wrote James Reid, a chronicler of the industry, in 1878. Even so,
the misunderstandings about the exact nature of the telegraph continued. One woman went into a telegraph office to wire the
sum of $11.76 to someone and then changed the amount to $12 because she said she was afraid that the loose change "might get
lost traveling over the wire."

B
UT EVEN WITH the introduction of security measures, there were still ways to make money by abusing the telegraph. In 1886,
forty years after the first attempts to use the telegraph to make money on the horses, an Englishman called Myers tried to
bribe an operator at the Exchange Telegraph Company's office at the Haymarket in London to delay the transmission of racing
results so that he could place bets on the winners. He was arrested, but when the case went to court it was found that the
only telegraph-related law he could be charged under related to damaging telegraphic apparatus, something of which he was
clearly not guilty. Delaying the mail was illegal, but delaying a telegram was not. The law was subsequently extended to make
it a crime to alter, delay, or disclose the contents of a telegram. Myers's case was never tried because he killed himself
with an overdose of laudanum. But it highlighted yet another example of technological progress outstripping the lawmakers.

T
HERE WERE, of course, instances where intercepting telegrams was regarded as acceptable—when it was governments doing the
intercepting, in the interests of national security. As a result, diplomats and spies routinely used codes and ciphers to
protect their messages from the prying eyes of enemy governments, but with varying degrees of success. Perhaps the most notorious
example of an intercepted telegram was the Panizzardi telegram, and its unfortunate consequences for Captain Alfred Dreyfus
of the War Ministry in Paris—an episode that became known as the Dreyfus Affair.

On October 15, 1894, Captain Dreyfus, an artillery officer, was summoned to a meeting and asked to take some dictation. Once
he had written a few words, his writing was compared with a newly discovered document written by a treacherous staff member
at the War Ministry who was passing information to the Germans. On the basis of his handwriting, Dreyfus was accused of being
the author of the incriminating document and was arrested on the spot for high treason.

Two weeks later the news was leaked, and a newspaper,
La Libre Parole,
reported that Dreyfus had been arrested for spying and was suspected of being in the pay of Germany or Italy. The resulting
outcry was to split French society into two factions: Dreyfusards (broadly, liberals who believed Dreyfus had been framed)
and anti-Dreyfu-sards (pro-military conservatives who thought he was guilty). Since Dreyfus was Jewish, the anti-Dreyfusards
were accused of anti-Semitism, and the polarization of political views that followed inflamed anti-Jewish feeling and divided
the country.

As tension mounted, the Italian military attache, Colonel Alessandro Panizzardi, sent a telegram to his chief in Rome to say
that as far as he knew, Dreyfus was not spying for them, though there was always the possibility that Dreyfus reported directly
to someone higher up the chain of command in Rome. With speculation rampant in the press, Panizzardi pressed his superiors
in Rome to issue an official statement if indeed Dreyfus was not one of its spies. Panizzardi sent a telegram to this effect
to Rome—a coded message that was to become one of the most infamous telegrams ever sent.

The message was sent in a numerical commercial code, where different groups of numbers represented different syllables, letters,
and common words. And like all diplomatic telegrams, it was immediately intercepted by the French Ministry of Posts and Telegraphs,
and a copy was sent to the Foreign Ministry so that the codebreakers at the Rureau du Chiffre (literally, office of code)
could work on it. (The French were, once again, ahead of the field; France was, at the time, the only country to have an official
military codebreaking effort.)

The codebreakers soon recognized the numerical groups. They were from a commercial code published a few months earlier by
an Italian codemaker, Paulo Bara-velli. The code used single digits to represent vowels and punctuation marks, double digits
for consonants and certain common verbs, triple digits for common syllables, and quadruple digits for key words. This system
made messages written in the Baravelli code easy to spot.

In fact, the codebreakers already knew the Baravelli code from earlier that year, when a torrent of telegrams had been exchanged
between the count of Turin, a nephew of the king of Italy, and Duchess Graziolo, a legendary Italian beauty staying in Paris.
The head of French army intelligence thought this had all the hallmarks of a spy communicating with his spymaster, so he ordered
the messages to be decoded. But nobody could make heads or tails of any of them, because they were all written in numbers.
Eventually, a French agent broke into the duchess's rooms and found a small, highly scented book: her Baravelli codebook.
The messages were soon decoded and were found to express nothing more than what one official de scribed as "simple, elemental,
natural feelings"—a correspondence between lovers, not spies. The Baravelli code thus became known to the Bureau du Chiffre.

However, like many commercial codes, the Baravelli code could be customized, to ensure that messages encoded with it could
not simply be read by anyone who had a copy. Each page contained one hundred words numbered from oo to 99. These were made
into four-digit groups by combining them with the number of the page on which they appeared. But each page also had a blank
where an alternative page number could be filled in. By renumbering the pages of two Baravelli codebooks in an identical fashion,
two people could then exchange messages with a fair amount of secrecy—since there are an astronomical number of ways in which
one hundred pages can be reordered. What's more, some pages had blanks where additional words could be filled in, words whose
meaning would be unknown to anyone who intercepted any message. When the codebreakers tried to read Panizzardi's telegram
using the original page numbers, it was gibberish; he was clearly using his own page numbers for added security.

However, since one word of the message—"DREYFUS"—was known, it wasn't all that hard to figure out some of the page number
substitutions, and eventually a partially decoded message was produced: "IF CAPTAIN DREYFUS HAD NOT HAD RELATIONS WITH YOU,
IT WOULD BE WISE FOR THE AMBASSADOR TO DENY IT OFFICIALLY." Only the meaning of the end of the message was uncertain; the
codebreakers' best guess was "OUR EMISSARY IS WARNED."