Louis XIV's Great Cipher is well-known by Bazeries' codebreaking and his hypothesis about the identity of the Man in the Iron Mask. The present article is an attempt to provide a broader perspective to codes and ciphers during the reign of Louis XIV.
(0) Ciphers used by ambassadors (1661/1675/1676) ... figures 1-99 with diacritics etc. (another article)
(1) Louvois' Code/Cipher (1676) ... entries up to 300
(1B) Cipher for Envoy to Bavaria (1679) ... entries up to at least 496
(2) Codes/Ciphers (1689) ... entries up to about 471; low numbers reserved for single letters
(3) Another Code/Cipher ... entries up to about 417; low numbers reserved for single letters
(4) Louis XIV's Code/Cipher (1690, 1693) ... entries up to at least 420; low numbers reserved for single letters
(4B) Louis XIV's Code/Cipher with Ambassador in Rome (1690) ... entries up to at least 535; undeciphered
(5) Louvois-Lauzun Code (1690) ... except for single "862", entries up to at least 478; Unbroken
(6) Code/Cipher (1691) Broken by Bazeries ... entries up to about 587
(7) An Unbroken (?) Code/Cipher for Catinat (1691) ... entries up to at least 366; Unbroken
(8) A Great Cipher (1690) ... entries up to at least 898
(8A) Two Codes for Peace Negotiation at Rijswijk (1696-1697) with Provision for Code Switching ... 500/378 entries
(8B) Louis XIV's Code for Marquis d'Usson and a Similar Code (April 1702) ... 540/560 entries
(9) An Unbroken Code of Catinat (September 1702) ... entries up to at least 464
(10) Code between Guelders and Rheinberg (November 1702) ... entries up to at least 383
Codes/Ciphers during the Reign of Louis XIV
Mémoires et correspondance du maréchal de Catinat, Tome II (1819) (Google), p.343-352, includes a code of October 1676, which was sent by Louvois, Minister of War, to a M. de Bonnais. While this is a relatively small code of 300 entries, it has its merit in its non-alphabetical arrangement of entries (i.e., it is a two-part code). The following is a part of its decoding table.
Generally, it assigns three numbers to single letters and two numbers to syllables (consonant+vowel). A few frequent words are assigned one to three numbers. Eleven numbers are reserved as nulls.
Louvois' instructions accompanying this code is no less interesting than the code itself. The code was to be kept sealed until further instructions were received from the King, unless the receiver was attacked by an enemy (at the time, France was at war with the Dutch, allied with the Imperialists (Wikipedia)), in which case the receiver was to report the news with the code. The packet was also to be opened when letters, encoded with a code different from the current one, were received from a place attacked by an enemy. In any case, when the receiver opened the packet, he was to report the reason to Louvois. Except for these specified cases, the receiver must not open the packet and must be ready to return it when Louvois required it at the end of the campaign, even if it would be given him again.
The instructions show the French court was aware that unnecessary use of a cipher would compromise its security.
The war with the Dutch ended in 1678 with the Treaty of Nijmegen. From 1688 to 1697, France was again at war (the War of the Grand Alliance, recently called the Nine Years' War by scholars).
An image of a letter of 15 November 1679 from Denis de La Haye-Vantelet, French envoy to the Bavarian court at Munich, is available on a website for France-Bavaria cooperation.
It has entries up to at least 496, probably representative of letters and syllables plus a few names/words.
The following is my provisional transcription.
The content is interesting in its unkind remarks about the late Elector of Bavaria (Wikipedia), who had died in May 1679 and whose mother was from the House of Austria.
The code used in a letter of 6 September 1689 from the Marquis of Bethune (a French emissary to Poland) to Cardinal D'Estree had entries at least up to "471 se", with low numbers up to about 64 being reserved for single letters. While the assignment of numbers are not alphabetical, words with the same initial letter tend to occur in the same row (i.e., the same ones digit). Possibly, the original cipher did not provide separate tables for encoding and decoding but only introduced some irregularity in a one-part code. The sign "+" represents "s".
A line placed below a figure indicates some imperfection or varied termination but these are occasionally neglected. For example, "335" is "Cardinal" but with an underline under "5", it may read "Cardinau" to form part of the plural form "Cardinaux". Similarly, "181 Pologne" may be used to form "Pologn-ois" and "177 France" to form "Franc-ois". "221 service" may read "servi" to form part of "servir." This feature identified by Wallis is clearly seen in the Great Cipher (1690) below.
A letter of 8 July 1689 by Callet de Teil, a judge of the Parlement sent to Poland, to the King of France used a code which is similar to but different from the one above. It had entries at least up to "471 fi". Numbers 14, 16, 18, 20, 24, 28, 30, 34, 36, 38, 40, 44, 46, 48, 50, and 52 are nulls. 10, θ, and κ delete the preceding number. 6', 8', and 9' repeat the preceding number. Symbols like ч ... ч delete everything in-between.
By arranging the elements in a table in which figures run vertically and code elements run horizontally, the regularity of the code can be seen. More over, it is clear that the Teil-Louis XIV Cipher of July 1689 has a very regular assignment, with proper names assigned in the last two and half columns. The Bethune-D'Estree Cipher of September 1689 starts the alphabetical arrangement in the middle of the table.
See another article for more details of these codes.
These were both broken by English mathematician John Wallis. Although more than 450 entries of these codes was superior to 300 for Louvois' code of 1676, these codes have weaknesses compared to the latter. In particular, lower numbers were reserved for single letters and probably Wallis first attacked sequences of such low numbers. Further, the arrangement of these two codes was not completely random but had some regularity in that entries with the same initial letter tend to appear in the same row.
John Wallis and the idea of a universal library has a photo of the last page of one French cipher Wallis deciphered. As with the above, it has symbols to "signify what is between is to be deleted"and a symbol to "delete that next before it." The arrangement and size are also similar to the above.
|371||381||391 temoi||401 tout||411 temps|
|374||384 ca||394 ce||404 ci||414 co|
|375 et||385||395||405||415 entre|
|376||386 France||396 faveur||406||416 faveurable|
|377 il||387 jour||397 interest||407 intention||417 imperieux|
|378 mariage||388||398 Moscovie||408||418|
|379 proposition||389 Prince||399 par||409 pouvoir||419|
Louis XIV's secret instructions to Castaignere, French ambassador in Constantinople, dated July-August 1690 printed in Paul Rycaut, History of the Turks (Google), p.453 ff, use a code with figures up to 420. The present author partially decoded the letters and reconstructed the code (see another article). It is similar to the above. In particular, lower numbers are reserved for single letters and bigrams etc. having the same initial letter have the same ones digit.
The same code is used in Louis XIV's dispatch of 9 June 1693 printed in David Kahn, The Codebreakers, p.168. (Actually, the present author used this as a clue to decode the 1690 encoded instructions.)
Some statistics of the 1690 letters in code may be useful.
In plaintext French, the most frequent letter is "e", followed by "a", "s", "i", "t", "n." In those letters encoded in this code, the most frequent code group is 53 (s, 48 times), followed by 202 (de, 43 times), 59 (u, 33 times), 98 (re, 30 times), 185 (la, 28 times), 38 (n, 28 times), 50 (r, 27 times), 14 (e, 25 times), 2 (a, 25 times), 3 (a, 23 times), 54 (s, 23 times), 167 (pa, 22 times), 51 (r, 21 times), 83 (en, 20 times). The relatively low rank of "e" can be explained by the fact that syllables such as "be", "ce", "de", ... are assigned their own code numbers.
Two-group sequences occuring 6 times are 119 59 (vo u), 298 80 (ti on), 3 59 (a u), 331 38 (co n), 59 54 (u s), 98 50 (re r) and those occurring 5 times are 14 51 (e r), 167 53 (pa s), 202 185 (de la), 59 51 (u r). (data)
Louis XIV's secret instructions to Charles d'Albert d'Ailly, Duke of Chaulnes (Wikipedia), French ambassador in Rome, dated 10 July 1690, use a code with figures up to 535. The letter is presented on a French website.
Chaulnes had been sent to Rome in 1689 upon the death of Pope Innocent XI (Wikipedia), who had been opposing Louis XVI's measures to control the episcopal sees in France (see Gallicanism and droit de régale in Wikipedia).
The letter, partly in cipher, has not been deciphered. In part, it reads:
Bigrams: 306 180 4x; 22 306 3x; 12 32 2x; 523 172 2x; 118 434 2x
Full Data: frequency, frequency (in descending order),
While I was writing this section, discussion of this code started in Klausis Krypto Kolumne.
Harry Thompson, The Man in the Iron Mask prints in the appendix a letter in a French cipher that John Wallis could not solve (see another article). It is one of the letters from Louvois to Lauzun in Ireland (27 May 1690), written about a month before the Battle of the Boyne. The following is a portion of it.
This has a number as high as "862" but, apart from this single instance, 478 is the highest number.
Wallis could not decipher this, probably because numbers for single letters were not assembled in lower numbers but scattered around the whole, as in the 1676 code above.
Seven letters from Louis XIV or Louvois to General Catinat were encoded with a two-part code with entries 1-587. The code was broken by Etienne Bazeries, the achievement leading to a hypothesis about the identity of the legendary Man in the Iron Mask (see another article).
As with the 1676 code, the arrangement is non-alphabetical and single letters are scattered among the whole. It assigns one (q) to seven (e, i/j, n, s, t) numbers to single letters and one to three numbers to syllables (consonant+vowel). Frequent words are assigned their own numbers. (But "un/une" is not given its own entry.)
Some statistics of a letter encoded in this code (Louis' XIV to Catinat, 14 September 1691, see another article) may be useful.
The most frequent numbers are 42 (de, 110 times), 22 (en, 107 times), 124 (les, 97 times), 125 (ne, 85 times), 145 (la, 80 times), 97 (de, 74 times), 34 (de, 66 times), 503 (pour, 63 times), 337 (la, 62 times), 117 (ra, 61 times), 9 (re, 61 times), 47 (que 60 times). The most frequent two-group sequences are those for "les en" (14x), "en ne" (12x), "mon arm" (12x), "que vous" (12x), "fi cu" (10x), "de la" (10x), "cet te" (10x), "don ne" (10x). (See statistics: (1) encoded text, (2) frequency (descending order), (3) two-group frequency, (4) two-group frequency sorted by frequency.)
Mémoires et correspondance du maréchal de Catinat (1819), p.283, also includes a coded letter from Feuquières (Wikipedia), dated "De Pignerol, ce.25 janvier 1691." This appears to be encoded with a different code from that of Louis XIV and Louvois.
Émile Burgaud, Étienne Bazeries, Le Masque de fer (1893) (see another article) claims that Bazeries (in the third person) deciphered this (p.272), though no information on the content is disclosed. (The reference to "Veillane", a place taken by the French (p.19), could have been drawn from Mémoires (1819) and p.57 says no one including "the author" (probably referring to Burgaud) had been able to read it.)
The highest number used is 366 (hence called a small cipher in Le Masque de fer). The most frequent numbers are 91 (12 times) and 44, 51, 78 (11 times each). Sequences occuring multiple times are "44 200", "151 310", "301 151" (3 times) and "5 44", "51 63", "51 81", "57 78", "60 52", "60 300", "63 135", "65 91", "78 44", "78 181", "81 91", "91 22", "99 84", "101 193", "101 229", "129 44", "150 301", "156 99", "157 91", "210 51", "229 255", "235 60", "255 243", "300 57", "310 235", "317 101" (2 times).
Wikipedia has an image of a part of a great cipher from 1690, which has a number as high as "898 le Baron de". As with the above, entries beginning with the same letter tend to have the same ones digit. Four numbers are assigned to vowels and "&" and two numbers are assigned to consonants and frequent syllables ("luy 848 239", "ma 868 298").
Further, the same number can indicate several variant forms. For example, "15" is "genera,l,ux", meaning either "genera", "general", "generaux"; "249" is "oc,casion", meaning "oc" or "occasion"; "822" is "quart,ier,s", meaning "quart", "quartier", "quartiers", "quarts". This characteristic is also seen in ciphers from the Napoleonic age.
Two extant codes from 1696-1697 are described in Jörg Ulbert, "Zur Verschlüsselung französischer Ministerialkorrespondenzen" in Geheime Post (2015). It was for use during negotiations for the Peace Treaty of Rijswijk (Wikipedia) (1697).
Figures are used to represent single letters, syllables, and words. One of the two codes had entries 1-500 and the other had entries 1-378.
These two codes were intended to be used alternatively in the same message. They are given identifying numbers "379" and "720", which can be used to indicate to the recipient which cipher is used first. See another article.
Lambeth Palace Library preserves French coded letters of April 1702 deciphered by the British (MS 930, no.9-10 and no.4).
This appears to have been written by a minister of the Sun King.
This code employs a very similar scheme to that of the code used in 1690 and 1693 (see above). Low numbers are for the most part reserved for single letters (ordered alphabetically). Numbers 110 and upwards must have been arranged in a table as shown in the image below, in which the code numbers run line by line, while the entries in each column are generally arranged alphabetically. This allows one and the same table on a sheet of paper to be used for both encoding and decoding, while introducing some irregularity to the arrangement.
From the placement of the sequences ca-ce-ci-co-cu and qua-que-qui-quo-quu, the original table may have ended with 539, to which some additions may have been made in the margin.
In addition to the numerical code, the letter employs several roman numerals (such as xxi, xviii, xlvii (undecoded) for some names).
The most frequent code groups in the letter are: 92 ("s", 123 x), 311 ("de", 82 x), 91 ("r", 79 x), 510 ("ce", 59 x), 516 ("que", 59 x), 247 ("re", 56 x), 30 ("e", 56 x), 22 ("a", 55 x), 184 ("la", 48 x), 190 ("au/av", 40 x) (data).
This is very similar to the code used in the letter of the day before. The top-left portion of the nomenclature (i.e., the code part) even seems to be identical between these. Except for 596(quelque), which may be an error for 546, the range of code numbers must have been 1-559. (See the placement of ca-ce-ci-co-cu and pa-pe-pi-po-pu in the image below.) The word "ennemi" (which provided Bazeries with a breakthrough) is given a code number "222."
The most frequent code groups in the letter are: 46 ("s", 50 x), 331 ("de", 44 x), 44 ("r", 34 x), 174 ("le", 31 x), 20 ("e", 26 x), 347 ("se", 23 x), 466 ("que", 20 x) (data).
The War of the Spanish Succession between France and Spain on one part and the grand alliance of England, the Dutch, and the Emprror on the other was about to officially begin by declaration of war in May 1702 (though hostitilies had already begun in 1701 in Italy).
Hanover and Celle (subdivisions of Brunswick-Lüneburg) were on the side of the allies, while Wolfenbüttel (the elder subdivision of Brunswick-Lüneburg) was ruled by brother dukes Anton Ulrich (Wikipedia) and Rudolph Augustus. Anton Ulrich allied himself with the French and planned to take the duchy of Celle. Hanover and Celle prevented this by a pre-emptive strike in the night of 19-20 March by marching into Wolfenbüttel (Ragnhild Hatton, George I, p.88-89).
The 4 April letter was written upon receiving letters of 21 and 23 March after the first report of the march on Wolfenbüttel from another source ("343" not decoded). The recipient appears to be a commander in some places which might be attacked (possibly Marquis d'Usson below). The letter assures the recipient the King's approval of his conduct and tells that the King would write to him his intensions separately.
William III of England, the architect of the grand alliance against France, had died on 19 March (Gregorian calender). The information was not yet confirmed in Versailles and the writer mentions that if the information was wrong, he would start a war at once and anyway the death might not change the plan of the allies.
The King's letter of 5 April is addressed to Marquis d'Usson, who commanded the Duke of Wolfenbüttel's troops etc. and resided at Wolfenbüttel with the character of a minister (Cole (ed.) (1733), Memoirs of Affairs of State Containing Letters (Google), p.411; the marquis is also mentioned in Vault below).
It was written in response to the marquis' letters of 19, 21, and 23 March about the operation of Celle and Hanover. The King promised relief of Wolfenbüttel in case of war He also instructed the marquis to urge the dukes of Wolfenbüttel to maintain the French alliance but indicated some concessions.
An undecoded letter of Marshal Catinat of 15 September 1702 is preserved in the library of Canton of Aargau, according to the 1 May 2016 entry of a cryptology blog Klausis Krypto Kolumne. It begins as follows.
The highest number is 464. The most frequent numbers are 50 (18 times), 28 (17 times), 13, 22 (14 times), 30 (12 times), 84 (11 times), and 14, 393 (10 times). The unbroken code of Catinat (7) appears to be different from this in view of these characteristics. Moreover, numbers 28 and 50 occur only once in the undecoded message of (7). None of the known codes above can solve this, either. The most frequent two-group sequences are: "41 269" (5 times), "13 101", "50 84" (4 times), "50 28", "82 30", "101 22", "183 284", "269 17", "284 13" (3 times each). Of these, all instances of "269" are in the sequence "41 269". The successive occurrences of two-digit numbers may suggest the code is something similar to the one described in (4) above, rather than a purely two-part code as broken by Bazeries. This hypothesis may be supported by the fact that no less than 253 groups out of 591 are lower than 100. (See statistics: (0) frequency, (1) frequency (descending order), (2) two-group frequency, (3) two-group frequency sorted by the second group, (4) preliminary contact analysis, (5) contact chart, (6) frequency histogram in comparison to (8B) above.)
The letter appears to be addressed to Beat Jakob II Zurlauben (Schmeh; Büsser (2008)). (Another Beat Jakob Zurlauben was lieutenant general in the service of France and the owner of the regiment Zurlauben (Wikipedia in German). In 1702, Lieutenant General Zurlauben was in Italy (Vault, p.725, 192, 231, etc.) and the regiment served in Flanders (ibid., p.482, 579, etc.).)
The letter is written at the camp of "Inglesheim", a name not found in today's map. Several possibilities are discussed in the blog, of which one is Ingenheim, which is tolerably consistent with the historical background below.
The letter was written in the last period of Catinat's active career.
Nocolas Catinat was made a marshal of France in 1693 during the War of the Grand Alliance. When the War of the Spanish Succession broke out in 1701, he was in command in North Italy but was transferred to Alsace in September 1701 (Frey, p.86, 479) and established a camp near Metz (Sturgill, p.18, 19).
In 1702, it was proposed that the French should cross the Rhine at Hüningen (now Huningue in France, near Basel) and make a junction with the Bavarian forces, the new ally of the French. Duc de Villars considered the combined forces should march on the right bank of the Danube in cooperation with the army in the Italian front but was frustrated by the cautious attitude of Catinat. (Vault, p.20-21)
In the meantime, Catinat had taken a position near Ingenheim (near Strassburg, between Saverne and Haguenau) on 17 August (Vault, p.363, 824) and decamped on 30 August (p.366). He attempted at a diversion in favor of Landau under a long siege by marching down the Rhine via Drusenheim to Roeschwoog (p.366, 370) but, when informed that Landau was taken on 9 September by the Imperial forces under the Margrave of Baden, he decamped on 11 September (p.372) and withdrew to a camp near Strassburg on 14 September, with the right near Schiltigheim (a few kilometers north of Strassburg) and the left at Eckbolsheim (p.373).
Almost at the same time as the fall of Landau, the Elector of Bavaria showed his hand and took Ulm by surprise (Churchill, p.637). But the communication between Bavaria and the French was made difficult by the vigilance of the Imperialists (Vault, p.374). (As an example, the Imperial minister in Switzerland armed his servants to catch messengers and couriers (Churchill, p.637-638, Vault, p.839, Sturgill, p.21).) In a despatch to the King on 14 September, Catinat could only report the seizure of Ulm by information of merchants (Vault, p.379). At this time, Beat Jakob II Zurlauben in Zug, Switzerland, was an intermediary for handling correspondence between the French ambassador in Solothurn, Switzerland, and Ricours or Ricous, French minister to the Elector of Bavaria, through hostile territories (Büsser, p.75).
Learning the fall of Laudau, the Elector was concerned with the security of his own territory and countermanded the march of a detachment to join with the French. However, his messenger to Catinat to inform of this change of plan was captured (Sturgill, p.20-21, cf. Churchill p.637-638). It was only on 20 or 22 September that an officer arrived at Catinat's camp and informed him of the Elector's order that his detachment should go no further than Stühlingen (Vault, p.374-375; Sturgill p.22). Worse still, although Catinat used cipher in his letters, the Elector did not and the intercepts revealed his plans to the allies (Vault, p.374).
Finally, Villars was given command of forces to cross the Rhine and won a victory at Friedlingen (near Basel) on 14 October against the Imperial forces under the Margrave that marched up the Rhine. Although the French and Bavarian forces could not yet effect a junction, this opened a path between the Elector on the Danube and the French on the Rhine (Lynn, p.276).
(By the way, the day before, Villars reported wasting seven hours with his secretary in trying in vain to decipher a letter he received from Ricous, which employed a small cipher different from the cipher between the Elector and Catinat, of which Villars had a copy (Vault, p.404, 414, 839, 841-842, 858; the note on p.841 says the latter part of the letter in the archives (Archives du dépôt de la guerre, original, vol.1582, no.97) is not deciphered).)
Catinat soon resigned for retirement (Frey, loc. cit.).
Klaus Schmeh, Wer knackt diesen verschlüsselten Brief von Nicolas de Catinat?, 1 May 2016, Klausis Krypto Kolumne (blog entry) ... images are available
Bernard (ed.) (1819), Mémoires et correspondance du Marechal de Catinat de la Fauconnerie, Tome III, Livre X (Google) This prints a letter from the Elector on 23 September 1702 on p.174.
Vault (ed.) (1836), Mémoires militaires relatifs à la succession d'Espagne sous Louis XIV, Tome II (Google)
Nathalie Büsser (2008), "Salpeter, Kupfer, Spitzeldienste und Stimmenkauf: die kriegswirtschaftlichen Tätigkeiten des Zuger Militärunternehmers und Magistraten Beat Jakob II. Zurlauben um 1700 für Frankreich", Schweizerische Gesellschaft für Wirtschafts- und Sozialgeschichte (pdf) ... see esp. p.75 and note 32
Linda Frey and Marsha Frey (ed.), The Treaties of the War of the Spanish Succession, An Historical and Critical Dictionary
Claude Sturgill, Marshal Villars
John A. Lynn, The Wars of Louis XIV 1667-1714
Winston Churchil, Life and Times (2 vols.), Book One, George G. Harrap & Co. Ltd. (London, 1947 ed.)
ViaMichelin ... Searchable map
Lambeth Palace Library preserves a French coded letter of 20 November 1702 deciphered by the British (MS 930, no.5, no.6). Addressed to Monsieur de Rogon, at Guelders, from Rheinberg, 20 November 1702, it conveyed instructions to a commander of dragoons.
Unlike the codes used by the King in April (see above), it employed two-part principles (i.e., random arrangement of entries). However, it had less than 400 entries (up to at least 383) and even the word "ennemi" had to be spelled as "190(en)-232(ne)-156(mi)" (twice). The British decipherer succeeded in reading it for the most part.
At the beginning of the War of the Spanish Succession, the French held all the major towns along the Meuse and the lower Rhine (with an important exception of Maastricht) and even the Dutch city of Nimwegen was threatened. However, the Duke of Marlborough commanding the allied forces completely turned the table. In September, shortly after the news of the capture of Landau from the French reached, Venloo fell. Roermond, Stevensweert, and, further up the Meuse from Maastricht, Liège were captured by the end of October. On the Rhine, Kaiserswerth had been captured from the French in June and Marlborough arranged for a winter siege of Rheinberg. Now, the control of the Meuse and the lower Rhine was in the allies' hand but the town of Guelders, fortified and protected by marshes, between the rivers was still held by the French (Churchill, p.606).
As of early November 1702, the allies went into winter quarters except for those besieging Rheinberg, which fell in February 1703 (Churchill, p.605, 604).
Numerical cipher without use of arbitrary symbols appears to have become the standard in France in the early 17th century (see another article). Thus, the codes/ciphers during the reign of Louis XIV used 300 to about 900 numbers to represent letters, syllables, and frequently used words and names.
Three diplomatic codes from 1661, 1675, and 1676 use two-digit figures with diacritics (see another article), the practice not seen in later ciphers known to me.
The 17th century saw introduction of two-part code, whereby entries are not arranged alphabetically and thus separate tables are provided for encoding and decoding. It is often said to have been devised by Antoine Rossignol (1600-1682), who served Louis XIII as well as Louis XIV as a codebreaker. He must have known from his own experience that codebreaking would be made more difficult by irregular arrangement of entries.
However, the non-alphabetical arrangement was not always fully implemented. For example, the above ciphers of (2) 1689 and (4) 1690-1693 had single letters in alphabetical order in low numbers. Moreover, the arrangement of syllables and words was not completely random but those with the same initial letter tended to follow one another with an interval of ten. Probably, the original cipher was arranged on a sheet as in the images shown above, which allows the single sheet to serve for both encoding and decoding purposes. (So, technically, it is a "one part" code.) The regularity allowed John Wallis, an English mathematician, who deciphered many French ciphers during the English Civil War as well as after the Glorious Revolution, to guess that, for example, a certain code number indicated a name beginning with a D.
On the other hand, as early as (1) 1676, Louvois had a cipher that appears to adopt a completely random arrangement of not only words but also single letters, both mixed together. Indeed, Wallis could not decipher Louvois' dispatches of (5) 1690, which is probably based on such a random arrangement, and others of (6) 1691 were only broken at the end of the 19th century.
In addition to common techniques such as homophonic substitution (having more than one symbol) and nulls (insignificant symbols), Wallis' deciphering of the (2) 1689 cipher and (3) another cipher shows that the French also used a symbol to delete a preceding symbol and another symbol used in a pair to delete everything in-between.
Use of a symbol to represent several variants (as in those of (2) 1689 and (8) 1690) is a feature that would be also seen in French ciphers in the Napoleonic age. While there should be a line placed below a figure for disambiguation, these were occasionally neglected, according to Wallis.
Wallis observed in 1689 the French "change their cyphers so often, and study every time to make them harder than before". In 1697, he said in his autobiography "of late years, the French Methods of Cipher are grown so intricate beyond what it was wont to be, that I have failed of many; tho' I have master'd divers of them." (see another article) Notwithstanding his testimony, later ciphers were not necessarily better than previous ones, as noted above.
Why did Louis XIV use a weaker one-part code as late as (8B) 1702 and (4) 1690/1693, when a securer two-part code was known at least as early as (1) 1676. I was made aware of an explanation by comments by Norbert in Klausis Krypto Kolumne. The two-part codes of (1) 1676, (6) 1691 (the one broken by Bazerie), and possibly (5) 1690 concern Louvois (Wikipedia), Minister of War in 1662-1691. On the other hand, one part codes such as (2) 1689, (4) 1690 were used by diplomats, presumably under Colbert de Croissy, Foreign Secretary (Wikipedia). Rivalry between the ministers may have prevented the principle of two-part code from being shared with ambassadors.
Codes/ciphers tend to be preserved by those who broke them. Inspection of John Wallis' papers would be desirable to obtain further insights on French ciphers during this period.