Tours, 1834 to 1836 · the Chappe telegraph · a trial record, replayed

The Error Was the Message

From August 1834 to August 1836 two Bordeaux bankers, the twins Louis-Joseph and Louis-François Blanc, knew the Paris price of the 3 per cent rente a day before the mail coach brought it. Two operators of the government's optical telegraph at Tours slipped one false signal into official dispatches and cancelled it at once with the regulation sign 'erreur', so nobody who could read a dispatch ever saw it, while anyone watching the arms at Bordeaux could. Plant a signal on the line yourself, read the trial record, then run the gloves-and-stockings rule their Paris agent was given over 602 days of prices transcribed from the Journal des débats: it sends 125 to 136 parcels, where the mail loader counted 121.

In the 1830s France already had a national telecommunications network, and it belonged to the government. Towers stood on hilltops and church roofs within telescope range of each other; one line ran from Paris by way of Tours and Bordeaux to Bayonne, and Du Camp counts a hundred and eleven stations on it [3]. On each one a long beam (the régulateur) carried a shorter arm at either end (the indicateurs), and the three could be set in dozens of distinct poses. An operator watched the tower upstream through a telescope, set his own arms to the same pose, and waited until the tower downstream had copied it. A dispatch crossed the country as a relay of shapes.

The men who worked the arms were not supposed to know what the shapes meant. Maxime Du Camp, writing about the service in 1867, put it flatly:

« Les préposés ignoraient absolument la valeur des signes qu’ils employaient. »

The operators had no idea at all what the signs they used meant.

Maxime Du Camp, "Le Télégraphe et l'administration télégraphique", Revue des Deux Mondes, 1867 [3]

Only a few officials along a line could read a dispatch. That arrangement kept the government's news secret, and it is exactly what two bankers in Bordeaux found a way through. The prosecution's own account of how they did it, published in the Gazette des Tribunaux on 28 January 1837, turns on one regulation sign of the telegraph code, the one called erreur.

Where a signal survives

This is the whole mechanism, in the indictment's words (read from the page image, not from its OCR):

« L'emploi de la malle de Paris jusqu'à Tours, pour faire parvenir le cours des fonds, de préférence au télégraphe, qui servait ensuite de Tours à Bordeaux, s'explique par l'existence d'un directeur à Tours. Ce fonctionnaire, investi du pouvoir de lire les dépêches, les corrige à leur arrivée dans cette ville et les purge des faux signaux qui peuvent s'y être glissés, avant de les transmettre à Bordeaux. Ainsi, un signal de convention, parti de Paris, se serait nécessairement arrêté à Tours, tandis qu'à Poitiers et Angoulême, les simples inspecteurs qui s'y trouvent ne pouvant lire les dépêches, ni par conséquent les corriger; un faux signal fait à Tours devait être répété sur la ligne jusqu'à Bordeaux. »

Using the Paris mail coach as far as Tours to carry the price, rather than the telegraph, which was then used from Tours to Bordeaux, is explained by there being a director at Tours. That official, empowered to read the dispatches, corrects them on their arrival in the town and purges them of any false signals that may have slipped in, before passing them on to Bordeaux. So an agreed signal sent from Paris would necessarily have stopped at Tours, whereas at Poitiers and Angoulême the mere inspectors there cannot read the dispatches and so cannot correct them; a false signal made at Tours had to be repeated along the line all the way to Bordeaux.

Acte d'accusation, extract, Gazette des Tribunaux, 28 January 1837, p. 313 [1]

« Maintenant lorsque Guibout, d'après les paquets reçus de Paris, voulait faire passer la cote de la Bourse à Bordeaux, il introduisait un faux signal dans une dépêche qu'il avait à transmettre; il le corrigeait ensuite à l'aide d'un signe réglementaire nommé erreur. Ce signe est employé pour indiquer que l'on s'est trompé et que le signe précédent doit être considéré comme non avenu; or, d'après ce qui vient d'être dit, ce faux signal et le signe erreur qui le suivait, arrivant jusqu'à Bordeaux, rien n'était plus facile au correspondant, au fait de cette manœuvre et qui observait les télégraphes de cette ville, de recueillir le signe erroné et de le transmettre ensuite à qui de droit, c'est-à-dire aux frères Blanc. »

Now, when Guibout, going by the parcels received from Paris, wanted to pass the Bourse price to Bordeaux, he put a false signal into a dispatch he had to transmit, then corrected it with a regulation sign called erreur. This sign is used to show that one has made a mistake and that the previous sign is to be taken as not sent. Since the false signal and the erreur after it went all the way to Bordeaux, nothing was easier for the correspondent, who was in on the trick and watched the telegraphs of that city, than to collect the erroneous sign and pass it on to whom it concerned, that is, to the Blanc brothers.

the same page [1]

Put a false signal on the line yourself. The line below is schematic: the real one had many more posts, and only the roles the indictment names are drawn. A blue square is a station where someone can read the dispatch; a gold circle is an inspector who, per the indictment, cannot; the rest copy shapes. Each row shows what that tower displayed, which is what a telescope trained on it would have recorded.

Try the three that matter. Planted before Tours, the pair reaches a man who can read, and he re-sends the dispatch clean: the signal dies there, which is why it could not simply be sent from Paris. Planted at Tours post no. 4, after the director has read, it rides every tower to Bordeaux, where the director reads the dispatch exactly as Paris wrote it, because erreur told him to. And planted without erreur, it reaches Bordeaux as a word that is not in the dispatch: a garbled message, which (our inference, not the record's) is the kind of thing that gets asked about.

So the correction is the trick. The one sign in the code whose job was to keep a dispatch true is what made the false signal invisible to every person able to read the dispatch, and plainly visible to anyone at the far end who could only watch the arms. It is a covert channel built entirely out of the error-handling.

Gloves for a rise, stockings for a fall

The price still had to reach Tours, and it went by the ordinary mail coach, in a parcel. The indictment gives the instructions the Blancs' Paris agent, Henri-Christophe Franck, a former non-commissioned officer and pensioner of the Invalides, received on 6 April 1835 when he took over the job from a man named Gormand:

« Lorsque sur le 3 pour cent, fin courant, il y aurait une hausse de 25 centimes, Franck devait envoyer des gants à un nommé Guibout, fabricant de bas, rue Montfumier, 27, à Tours. S'il y avait une baisse de 25 c., il fallait envoyer des bas ou des cravates. Franck devait ajouter 25 c. au cours du dernier jour du mois, pour pouvoir calculer, le lendemain, la différence entre le cours du jour et celui de la veille, à cause de la liquidation. Le 7 juin et le 7 décembre il ajoutait 1 fr. 50 cent. au cours et s'il y avait après cette addition une différence de 25 cent., avec le cours de la veille, en hausse ou en baisse, Franck envoyait, suivant le cas, des gants ou des bas, etc. »

When the 3 per cent, for end-of-month settlement, rose by 25 centimes, Franck was to send gloves to one Guibout, stocking maker, 27 rue Montfumier, Tours. If it fell by 25 c., he was to send stockings or cravats. He was to add 25 c. to the price on the last day of the month so as to work out, the next day, the difference between that day's price and the previous day's, because of the settlement. On 7 June and 7 December he added 1 fr. 50 to the price, and if after that addition there was a difference of 25 c. from the previous day's price, up or down, he sent gloves or stockings accordingly.

Acte d'accusation, Gazette des Tribunaux, 28 January 1837, p. 312 [1]

That is an algorithm, and a careful one. Two of its clauses correct the price series for jumps that carry no news. The indictment gives no reason for the 1 fr. 50 beyond the dates, but the Journal des débats supplies one: it is the half-year's interest on a 3 per cent rente of 100 francs nominal (3 ÷ 2 = 1.50), and its Bourse report for Monday 8 June 1835, the first sitting after Sunday the 7th, opens with "Le détachement du coupon de la Rente Trois pour cent s'est fait à 77 fr. 60 c." and reports the 3 per cent fin du mois down 90 c., "coupon détaché" [8]. Without the addition, the coupon coming off would have looked like a fall. The 25 c. at the month's end is given only as "because of the settlement"; it looks like the step between one month's settlement price and the next. Neither correction would matter unless someone meant to act on a difference of a quarter of a franc.

Two statements drawn up by the man who loaded the Paris mail, Chauliac, and certified by an inspector of posts, counted the parcels:

« … depuis le 22 août 1834 jusqu'au 25 août 1836, 121 paquets ont été envoyés à Guibout, soit par Gormand, soit par Franck. Suivant Chauliac, ces paquets toujours ouverts en sa présence, avant d'être expédiés, ne contenaient que des gants ou des bas … »

… from 22 August 1834 to 25 August 1836, 121 parcels were sent to Guibout, either by Gormand or by Franck. According to Chauliac, these parcels, always opened in his presence before being sent, held nothing but gloves or stockings …

Acte d'accusation, Gazette des Tribunaux, 28 January 1837, p. 312 [1]

What the rule would have sent

Franck's rule is specific enough to run. The Journal des débats printed every day's Bourse in a table, and the row it calls 3 p. cent, Fin courant is the quotation the rule names, with its first, highest, lowest and last price. We transcribed that row for every Bourse day from 14 August 1834 to 30 August 1836: 617 issues, 2,261 readings in 19 separate reading passes, every crop read at least twice by readers who could not see each other's work. A day's four prices were accepted only where two readings agreed exactly. That gave 588 of 602 Bourse days. Thirteen stayed disputed and are left empty; 22 are kept exactly as the paper printed them although a first or last price falls outside the day's own high and low; one impossible figure ("50 55" on a day in the 80s) is set aside, not corrected. How the crops were cut, read and reconciled, with every Gallica address, is in the research notes. The rule then runs with the indictment's constants and three stated reading choices: "a rise of 25 centimes" means at least 25; the month-end 25 c. goes on the last Bourse day of each month and the 1 fr. 50 on the first Bourse day on or after 7 June or 7 December; and the parcels stop from 30 April 1835 until a day in August that the record does not give.

With the last price of each day, the rule sends 136 parcels over the 494 day-to-day comparisons the transcription supports if the pause ended on 31 July 1835, 131 if it ended on 15 August, and 125 if on 31 August. Chauliac counted 121. Run on the prices the Paris Bourse actually made, the rule as the indictment states it produces the loader's count to within 3 to 12 per cent, slightly over.

The threshold is the part this tests hardest, because the count is steep in it: 191 parcels at 20 centimes and 94 at 30 (last price, pause to 31 July). The loader's 121 falls between those two, at the value the indictment gives. Nothing here was tuned to reach 121; the threshold, the two additions and the window are all the indictment's.

What it does not settle. Which price Franck compared: the first-price replay has only 352 usable comparisons, because many crops cut the first column, and on those it sends 113. Whether every parcel the rule called for was sent, or what the colour instructions of August 1835 changed. The rule runs a little high on every reading we tried, and the record does not say why. Two robustness checks: on the 213 comparisons where both agreeing readers were sure of every digit, the rule fires on 62 (29 per cent of days, against 136 of 494, 28 per cent, overall); leaving out the 72 rows whose date comes from the issue rather than a legible heading, it fires on 103 of 378 (28 per cent).

Someone else was reading

A channel anyone can see is a channel anyone can read. Among the papers seized from Guibout on 18 August 1836 was a letter dated from Ax, in the Ariège, on 15 July 1836, which the prosecution attributed to the Blancs (they were staying at the Hôtel d'Espagne there that month, and handwriting experts said several of the letters were theirs):

« J'ai envoyé Renaud sur la route de Paris, pour voir s'il y avait un service de pigeons, et il n'y en avait pas. J'ai su qu'à Bordeaux quelqu'un recevait le cours de la rente comme moi; puisque ce n'est pas par des pigeons, c'est donc par le même moyen que moi. Il est facile de voir à Tours quelles personnes expédient de semblables dépêches. L'individu qui les exploite à Bordeaux, n'ayant pas l'habitude, n'a pas su opérer comme il fallait. S'il reçoit encore 5 ou 6 dépêches, l'affaire sera tout-à-fait perdue, car on aura la certitude que l'on connaît le cours un jour plus tôt. »

I sent Renaud along the Paris road to see whether there was a pigeon service, and there was not. I have learned that in Bordeaux someone was getting the price of the rente as I do; since it is not by pigeons, it must be by the same means as mine. It is easy to find out at Tours who is sending such dispatches. The person exploiting them in Bordeaux, not being used to it, has not known how to operate properly. If he gets 5 or 6 more dispatches, the business will be entirely lost, because it will be certain that someone knows the price a day early.

letter from Ax, 15 July 1836, as quoted in the acte d'accusation, Gazette des Tribunaux, 28 January 1837, p. 313 [1]

Guibout's draft reply, dated 24 July, insisted that "il est impossible qu'il y ait à Tours quelqu'un qui fasse les mêmes affaires que moi" (it is impossible that anyone at Tours is doing the same business as me). The record does not settle who the other reader was. The letter does settle what the Blancs themselves thought they were buying: the Paris price a day early, and a secret that would not survive anyone noticing a pattern. By then the pattern was already being noticed. The Gazette's report of 10 December 1836 says the administration had been told in 1836 "de nombreuses erreurs commises exprès" (of many errors made on purpose) [2].

Not an act of his office

The case was tried before the Cour d'assises of Indre-et-Loire at Tours from 11 to 14 March 1837. On the bench were Pierre Guibout, 34, the operator; Louis-Joseph and Louis-François Blanc, 30, twins, bankers in Bordeaux; Guibout's wife Zélie Morion; and Pierre Renaud, 31, a former telegraph operator from Lyon who had received the signals at Bordeaux. Guibout's colleague at post no. 4, Lucas, was dead: the indictment gives 21 January 1836, and on his deathbed he had told a friend named Cailteau about the arrangement, one of the things that reached the administration in 1836 along with the director's reports of rumours [1, 6]. The charge was corruption of a public employee, under articles 177 and 179 of the Penal Code [2]. The first three questions put to the jury were about Guibout:

Questions and answers as printed in the Gazette des Tribunaux of 17 March 1837 [4]; our translation. The jury answered the questions on the Blancs' corrupting Guibout, Lucas and a third man, Devaux, "oui, mais sans la circonstance énoncée à la deuxième question", and no to everything else.

The facts were found; the crime was not. Bribery under article 177 needed the bribe to buy an act of the official's function, and the jury said sending false signals was no part of Guibout's function. The court acquitted Renaud and Zélie Morion, and ordered Guibout and both Blancs to pay the whole cost of the proceedings, granting the prosecutor's request made "attendu qu'ils ont par leur conduite illicite amené le procès" (since their unlawful conduct had brought the case about) [4]. Du Camp's gloss thirty years later was that they were "non pas acquittés, mais absous", not acquitted but absolved [3].

The government had not waited for the verdict. Du Camp says the investigation itself had opened the ministry's eyes, and the minister of the interior, Gasparin, introduced a bill in January 1837, before the trial [3]. It became the law of 2 May 1837, a single article, whose first paragraph reads:

« Quiconque transmettra sans autorisation des signaux d'un lieu à un autre, soit à l'aide de machines télégraphiques, soit par tout autre moyen, sera puni d'un emprisonnement d'un mois à un an, et d'une amende de 1,000 à 10,000 fr. »

Whoever transmits signals from one place to another without authorisation, whether by telegraph machines or by any other means, shall be punished by one month to a year in prison and a fine of 1,000 to 10,000 francs.

Loi du 2 mai 1837, as reprinted in the Almanach national (1891); we read it in a Google Books snippet [5]

It does not mention the Bourse, or bribery, or the telegraph administration's own staff. It makes the signalling itself the offence, by anyone, by any means, which is the state claiming the whole medium. The trial record shows why "any other means" was not a flourish: the Blancs told the court that speculators ran pigeon services and private telegraphs everywhere, and the administration's own witness, Allard, described a chain of windmills near Angoulême that passed a rise or fall down the road to Bordeaux by the set of their sails, and a festival day on which a window left open by accident sent the wrong one [6].

What the retellings get wrong

This story is often told as "the first cyberattack" [9], and the retellings have drifted. Set against the trial record:

• The director at Tours was not the one bribed, as the French Wikipedia article on the affair has it. He was the obstacle. The director, Bourgoing, is the man who reported the rumours about Guibout in May and June 1836 [6]; the whole design exists to get past him.
• "Acquitted because there was no law against it" is half right. No law made private signalling a crime, which is why the law of 1837 was needed, and Du Camp says the case "n’avait pas été prévu par la loi" [3]. But the charge was bribery, and the jury found the signalling and the bribes proven; it answered only that sending the signals was not an act of Guibout's office, which the bribery article required [4]. The conduct was not found lawful. The court made them pay for the trial.
• The deathbed confession was Lucas's. Lucas, the second operator, told his friend Cailteau as he was dying [1]. Guibout confessed too, but only at the trial [6].
• 121 is a count of parcels, not of false signals. Du Camp wrote that the brothers "avaient reçu cent vingt et une fois le faux signal" [3]; the indictment's 121 is the mail loader's count of parcels sent to Guibout [1]. They may be the same number, but the record counts parcels.
• The man watching at Bordeaux was, in the trial record, Renaud, who testified he was "chargé par les frères Blanc de recevoir les signaux du télégraphe à Bordeaux" [6]. Du Camp instead has the operator of the Saint-Michel tower pass the signal to the Blancs' clerk [3]. The contemporary record is the better witness.

Dumas was paying attention. In The Count of Monte Cristo, first serialised in 1844, the count bribes an operator on a line south of Paris to send a false dispatch, "Le roi don Carlos a échappé à la surveillance qu’on exerçait sur lui à Bourges", and Danglars, selling on it, loses half a million francs (chapter 61) [7]. The count's remark to the operator is the premise of the whole affair: "vous répétiez des signaux que vous ne compreniez pas vous-même", you repeat signals you do not yourself understand.

Sources

1. Gazette des Tribunaux, 28 January 1837, "Cour d'assises d'Indre-et-Loire. Affaire des télégraphes", extract of the acte d'accusation, pp. 312 to 313. scan (ENAP).
2. Gazette des Tribunaux, 10 December 1836, report of the indictment by the Cour royale d'Orléans of 6 December 1836. scan (ENAP).
3. Maxime Du Camp, "Le Télégraphe et l'administration télégraphique", Revue des Deux Mondes, 2e période, tome 68, 1867, pp. 457 to 497. Wikisource.
4. Gazette des Tribunaux, 17 March 1837, hearing of 14 March: questions to the jury and verdict. scan (ENAP).
5. Loi du 2 mai 1837 sur les lignes télégraphiques. We read it only in Google Books snippets, not a page image: the Almanach national (1891, id 6YpVAAAAYAAJ) and Das Telephonrecht (1885, id RWxy4N_dg4wC) give the wording quoted; the Recueil des décisions du Conseil d'État (1852, id T3ZDAQAAMAAJ) calls it "la loi du 2 mai 1837, dont l'article unique est ainsi conçu"; Dalloz's Répertoire (1861) restates it in the present tense.
6. Gazette des Tribunaux, 13 and 14 March 1837 (one issue), hearing of 11 March: interrogations of Guibout and the Blancs, testimony of Allard, Bourgoing, Franck and Chevreuil. scan (ENAP).
7. Alexandre Dumas, Le Comte de Monte-Cristo, chapter 61, "Le moyen de délivrer un jardinier des loirs qui mangent ses pêches" (chapter 60 is "Le télégraphe"). Wikisource.
8. Journal des débats politiques et littéraires, daily Bourse tables, August 1834 to August 1836, Bibliothèque nationale de France, Gallica (ark:/12148/cb39294634r). Each day's crop is listed with its Gallica address in data/prices.json. The coupon report quoted above is the issue Gallica dates 8 June 1835 (ark:/12148/bpt6k438786x), page 4, "Bourse de Paris du 8 juin".
9. For the "first cyberattack" framing: Bruce Schneier, "1834: The First Cyberattack", Schneier on Security, May 2018, quoting Tom Standage in 1843 magazine. schneier.com.