Doc ID 6618711 Experiences l 920-1939 l lY 8RIGADJ - H JOHN H TILTMA 'l 11$3tC li t ti JQ ftc ent s cry ptanalytic experience of Brigadier John H Tillman in India and EnRland durinR the period 1920- 1939 both a - Briti h Army offi cer and a War Office civilian crypiana ly t Di cu e o detail of some early Soviet transposition additiv e and one- time pad ystem and de cribe th e role of everal other Briti h military and civilian cryptanalys s Durin I ht- sumrn r of 1 920 I was on a Russ ian language 1 uurse in Lo11du11 Al the ud uf il I was about lo return to regimental duty but the War Office intervened and on l August I went to work on temporary attachment for 2 weeks at the Cmoemment Code and Cipher School wh ich at the time had a growing backlog of untranslated Ru ssian dip lomatic messages After n few days the War Office decided to hold me there for a year and in foct I never returned to my regiment When I joined it the Government Code and Cipher School occupied the whole of Waterga te House on the Thames Embankment near Charing Cross GC CS was formed after World War I of offi cers from Admiralty 40 OB and the Wa r Office Cork S treet office It was directed by Commander A C Denniston who had been in 40 OB during the war His deputy was Commander E V Tra vis lalt r Sir Edward Travi s our director 1942-195 2 who lil the J ime repre p nJ eri J hP Admiralty on COMSF rna1ters I worked as one of a group of from 5 to 7 perso ns on Russia n d iplomatic ciphers under the direction of Ernst Fetterlein Fetterlein had been Chief Cryptanalyst of the Ru sian Czari t overnment and held t he ranks of both adm ira l and general he had pract iced cryptanolysis since 1898 or earlier At the Revolution he walked out of Russia across the Finnish frontier and waH speciall_v naturalized on arrival in En land At the time of my arrival Fetterlein's s mall section was entirely occupiEd with the solution or the current Moscow-London and LondunMosenw diplom a tie rra ftii inl ereepted in I he cab le oflice All messages were enciphered by simple columnar transposition of Russian plain text conventional transliterated ou t of Cyrillic charactel'5 As each message was trans posed on a different key all messages had to be individually solved The a vcragc delay was I believe l or 2 diws nrmr1t1Dfllltlflflr1m1mmnrimmrmn11u11111mnu 1111111 HI 5 Jtt l di ESRA Approved for Release by NSA on 05-15-2018 FOIA Case #102197 Ill Doc ID 6618711 TCP EXPERIENCJo' S About ovember 1920 the system changed dinomes being substituted for the plaintext letteN before transposition The substitution table used provided variant dinomes for the letters according to frt quenl'y tht rn being I rl'ml'mber 7 variants for t'itl'h of thl' vowels In tran position thl' dinomes were not plit o Thl' original dinmnic sub tit ut ion table was nlved hy Fetterlein from 11 hust''-one of the tr11 r1spo ition keys of the e11rlier system was reused Solution of the individual messages then proceeded more or le as before But early in 1921 a new substitution table was introduced and messages from that time carried the cipher di criminant DELEGAT in the preamble There were this time no busts and for some weeks no progress was made The only hope appeared to be to find a message containing a long Hush repetition and this could only occur if the variants were badly used I made the first entry about April 1921 I mm 1 admit that I wns very lucky in finding an unusually favorable message None of the workings of those days have urvived and I have had to make up a simplified example to show what happened Suppose Table A to be the substitution table in use The message which gave me the solution contained the word DOGOWOR meaning treaty seven times In each case the encipherer had used the first Second and third variants for the letter O in that order and in fact had produced an identical 14-digit sequence 11 07 04 18 17 22 19 seven times Two of theSEUR se4mm es started in the same column of tht' transposition matrix a can he seen i11 the skelt'ton e1 ample Tahle R in whicb all dinnrne have been omitted except tho e forming part of the 14-digit sequence representing DOGOWOR and those entering into the beginning and end of the message The cipher mesMge is as - umed to have been written out in columns of 14 dinomes in order to show that if the keylength is 24 the elements of the sequence appearing twice starting in the same column will lie in the fourth and twelfth lines in the cage rc spcctively The mctisage was the second part of a three-part message and therefore the plain text began PHOU i e abbreviation for contir uation and ended PROD SLED i e continuation follows and these overlapped the flush repetition to a certain extent o Table C shows again in skeleton form the appearance of the cipher text on the assumption of a key-width of 24 The element of the flush sequences are squared In Table D the columns containinl the elements of the flush repetition were grouped together and rearranged when it wa s realized that the DOGOWOR sequence occurred a third time at an offset of one dinome from the start of the flush repetition oOn another li 1k a 1mil r system was u cd in which the dinomci were split ond columns or ingle rli2it' nan fK t f'rl In 1his system FNtE-rlr-m harl nmr surf'f' -l h11t onl- in 1 he CllllP S of one bt1 t and TOR 6 C l P Ii few perfect rectangle 2 Hlftlllllllll1111Hllllllll11JIJIIJllllllJ JDIJIIQIJIIJIQIIQIIII J H TIL'l'MAN liiP iliiFH l F From this point it was possible to complete the matrix as shown in Table E and recover a considerable part of the substitution table This enabled us tu reeuver 1 number of other tran pnsition kevs It had lnng been su pect d that t he keys were derived in the orthod x manner frum running tut probably from lines of poetry in view of the variation of key-length and Fetterlein had tried to reduce the recovered ke_vs o Hussian without ucces One da v I tried English instead of Russian and met with immediate success Fetterlein had shown me how to display a key by writing the consecutive numbers trom left to right and dropping a line every time a number was to the left of the next lower number Table F shows the derivation of the kev I recovered from the DOGOWOR message To cut a long story short 1 traced the source of the keys in the British Museum Library It turned out to be an out-of-print pocket edition of the works of George Wither an obscure poet of the Seventeenth century I do not remember the method of indicating keys but I know it was simple and that after finding the source book we were in a position to decr 'Pt DELEGAT messa es as soon as the intercepts reached us In the summer of 1921 the War Office was lookin for a replacement in India for Colonel W H Jeffery who had been an officer in a southern infantry regiment of the Indian Army but had in fact never served wit h hi f'e i111e111 He harl spent the first fow vear uf the Twenlielh Century in China 1Parning the langnagP and had i onP traight from there to South Africa in cha e of Chinese coolies in the mines About 1912 he had been posted to the intelli11ence branch M03 of the General Staff in Simla Simla wa the summer capita of lndia in the foothills of the Himalayas to which the Viceroy Commander in Chief and Governor of the Punjab repaired with their staffs from the plain rn the summer months There Colonel Jeffery took up the studv of Chinese ciphers To the best oi mv knowledge he reported only to the Indian Government and up till 1921 had not been personnlly in contact with GC CS or any other cryptanalytic group He had been very successful in reconstructing some Chinese ciphers consisting of very large one and two-part codes After World War I he had begun working on Russian ciphers He had been promoted to Brevet-Colonel and the General Staff had ruled that he could not retain his rank unless he returned to his regiment for duty This was a prospect he rould not face and he contemplated retirement as soon as a suitable rt'plaeernent could be found for him At fir t it was suggested th11t Captain A G S Muntz at that time working as a crypt rnal l t i11 Baghdad in haq-of Muntz more later should replace him and that I should go a asl'istant military attach to Meshed in Persia for which post 1untz was intended as some forward exploitation of Russian intercepts was contemplated there As a result of my success ur Ci SIM IDB SEC PEI I BIRR t EXPERlEKCES J H TILT AN with DELEGAT I was selected to relieve Colonel Jeffery and it was decided to drop the Meshed project and leave ' fontz in Baghdad After a month working as a cipher clerk in thi WHr 0 lice whic-h gave me experience for which I was later extremely grateful I departed for India in September 1921 While I was on the high seas the Indian General Staff relentl'd and dei i il'd to allow Colonel Jeffery to remain in Simla and retain his rank hut on the condition that he took a year' IP Hve of absence away from India leaving me in charge This ammgement -ie eyed with deep misgivings and we didn't part on the best of terms at the end of the year I remained eight and a half years in Sirnio staying there in the winter when mo t of the General Staff moved down to Delhi The Section MO3G consisted usually of Colonel Jeffery mvself one army officer Russian interpreter one clerk and one officer i nterpreter in Eastern languages chiefly Persian Our main a mO t our sole Russian taok was the exploitation of diplomatic messages passing between Moscow and Kabul in Afghanistan and between Moscow and Tashkent in Turkestan We had two radio intercept station civilian mannedL one at Cherat in the hills above Peshawar on the Nort hwe t frontier and one at Pishin in Baluchistan Llttl'r Vl' h 1d H slHtion al Maymyo in Burma fur a ti me When I first arrived the Russians were using dinomic syllabarics rec iphered by repeated cyclic use of 5-digit additives a different one as for as I remember for each message Thi was pretty easy stuff for Colonel Jeffery after his work on Chinese even although he refused absolutely to learn any H ussion Shortly after my arrival a new cipher with the discriminant name AZIYA was introduced This was lightly more troublesome as one of about seventy 20-digit additives was employed for each message first reading from left to right then from right to left then left to right and so on This was my first success in Simla dinome cut in groups of odd position but on it in groups of even position Therefore the 4th and 5th dinomes of each set of 10 digits retained the frequencies of the origirial syllabary Table G shows the effect of reciphering a common clichEUR such as AFGA -IPRA abbreviation for Af11han Government on the 5 possible cuts I reconstructed the KO 'IISlJL syllabary from the portions of mesSBges which retained t he dinome cut and read r lu mes sages wit hour fully urnlen tanding wfml lrnppenPu in the case 01 thP olhPr three-fift hs nf the lt xt Ry this time I was in touch with Muntz in Baghdad and had the mortification of being told the explanation by him I should perhaps say here that neither Colonel Jeffery nor I had any general cryptanalytic training nor any knowledge of statistics Our experience was limited to the very few types of systems which we had handled and there were no satisfactory technical books to which to refer Fetterlein had taught me a little but only in the field of transposition it used to be said in GC CS in 19 m 19 l that I was the only person Fetterlein hod ever been known to help The same idea in another form appeared a little later in the cipher ALTAI Here the intermediate text derived from a dinome syllabary was arranged in 6-figure groups and the be and de dinomes reciphered by use of a conversion table Table H shows the effect of applying the rules of ALTAI lo the element of Table G Sometime in the sprinl of 1922 I prophesied that the next change would be to continuous dinome conversions off the cut leaving only the first digit and the la digit of a Ilt'ssagl' unchanged I felt t xtremely ratified when very shortly afterwards this exal'l l'hange took phu e but HI first could make no headway This was partly due to the fact that the me age of different days of the month appeared not to repeat into one another Muntz produced an elaborate theoretical analysi and mE'thocl of solution based on the observation that repetitions in the intermediate text would be mostly represented in the c-ipher text by repetitions two digits shorter lt should therefore be pOS 1ible to group together al dinomes of the cipher text representing dinomes begmning with the same digit and b dinomes representing those ending with the same digit and thus produce a provisional table which would work as well as the true one But we both found in practice that this process produced only confusion with every dinome equal to every other So I took off to see Muntz in Baghdad carrying my problem with me After ahout three wel'k I olved I he trallic of nne day l y reading probable word such a AFGA 'IPRA into the repetitions in the cipher text and rernnstructing the yllabary and conversion table bit by bit It turned out that there were six different syllabaries and ix different conversion tables and these were u ' ed in various random combinations for the variou doys of the month TIP iiillif Doc ID 6618711 Colonel Jeffery departed on leave for England in December 1921 and I waE left in charge But just before he left the Russians had intro duced an entirely new type of cipher system Again I have to make up an example They used a dinome syllabary for the basic substitution and what we called a conversion table providing random din Jfrte toquivaler1ls for eHch of the dinornes 00 lhru 99 The intermediate text was reciphered hy use of the conversion table off the dinome cut leavin11 certain digits unaltered This svstem was introduced in three stages In the first system called KONSUL dinomes from the conversion table were substituted for the be and de dinomcs of the fi-tigure groups of the intermediate text leaving the first dii it of groups unaltered This produced the strange result that conversion was off the C SEC lti bl lbitA IPR SSCSEI t 102 SECPSI Doc ID 6618711 nr u EXPERJF NCES J H TTLTMAN Captain Muntz was an artillery captain younger than myself with a fine brain and an enormous capacity for work He was a good linguist with a knowledge of Russian entirely self-taught of about the same standard as mine I had just scraped through the Army language exc1111ination in 1920 as a second class interpreter I later passed first cl 1 s hut the standard was quite low as we couldn't go to Russia to learn We suffered tt l reat loss when Muntz died aged about 40 He spent a year or two at r C C S in 192 -1924 working in other linguistic fields and was consequent ly nnt with us when Colonel Jefferv and I were held up in our solution of Ru ian ciphers When I left india at the end of 1929 to start the Military Section ofGC CS Muntz tJ 1k over from me in Simla and remained there till he died Sometime in 192 1 the Russian introdul'ed long additives for thl' tirst time and in the absence of obviously significant repetitions in thl' cipher text and owing to our failure through lack of general experience to recognize concealed indicator3 we did not succeed in diagnosing the new type of cipher for several months From 192 l to 19'28 a succession of additive series all of the some general form were introduced applied to a number of code books widel_v ditfenng- in dimensions and form Starting point indicators were concealed I believe bv addition of textual groups in prearranged positions and placed in positions varying with the system The addith es were all 1000 5-figurc groups long arranged in 100 lines of 10 groups each and were applied 'boustrophl'rlon i e reading fir t line left to right then the next right to IPft P 1 c The starting point for a mcs -age wa cho er b_v the operator and could he at any of the 1000 groups of the additive The construction of the additive was fre 1uently far from random exhibiting various personal idiosyncrasies but was nol often predictable and had to be solved figure by figure During t he lifPti1111 of -rn a1ltlitive sufficient depth was usually provided for solution of a considerable portion of the traffic In the late summer and fall of 1924 I rei m en d a great part of the additive and the code book of the first of these svstems The code book contained 2 X l0 groups from 0000 to 1999 ani was completelv alphabetic i e one-part ln about November 1924 I went bark t England for a month vi iting on the way Baghdad a 1d our W1-1r Office intercept station at Sarafond in Pale tine It appeared that in London Fetterlein had in fact solved an additive system similar to ours and some time earlier than we realized what we had but the solution had not been reported to us in India Between the time when I returned to Simla and the summer of 1928 we recovered vast amounts of additive and t or 5 more code books I remember particularly one which contained a a aooo group 1-part rnde 4-figure l TOLIJlS beginning with 0 1 and 2 b -100 random trinomes repre enting the commonest words beginning with 3 4 5 and 6 and r a dinomic alphahPt the rtinomes 70 thru 99 representing thP let tprs of the Russian alphahet in alph hetic order Another code whi h did not retain a con tant cut contained 90 4-figure l rnup beginning 0 thru 8 and 10 000 figure groups beginning 9 ISP SEGPCT I 6 IUrn In the fall of 19 5 the Government of India sent a column known as the VA A column to the northwest frontier to occupy Wa2iristan to deal with unre t amon the northwest tribes a more stormv ituation than u ual Stark the Russian Ambassador in Afghanist n sent a cipher telegram to Moscow in which he inquired what joint action was proposed between the Russian and Afghan Governments 'in view of the occupation of Waziristan W Widu Okkupacii Waziristana Our interpreter who was quadriligual in Russian En lish French and Gt rman but not outstandint IY literalt' in any one of them tran h ted thi - with a view t o th oc e11p1Hinn of Wazirist an The intellip enc e hra11ch of Army Headquarters wn in Delhi and we were in Simla and there was a day of near crisis in Delhi before someone realizing- that it would take something like six months for Russians and Afghans to join force over the llindu Kush queried the translation back to us 1 well remember Colonel Jeffery Myinp 'Jn future all startling statements of this nature will be viewed with the utmoot suspicion The outcome was that I was told to check all our interpreter's translations before they went out My knowledge of Russian was very inferior to his but mv standards of accuracy were a lot higher and the new order meant a great deal of extra work for me At this time I was in fact involved in all aspects of the work dia11nosing the cipherg recovering the additives reconstructi lg the code books performing the rudimentary traffic analysi ta ks necessary visiting one of the intercept stations at Cherat and directing the intercept ccveraiie translating or checking tran lations and frequently having to argue the meaning of messagPs ith the general tafl' On two of my three visits to Baghdad I ' '' worked in the Baghdad interl'cpt station in the set room All this gave me breadth of experience which cumpc1ratively fi w other nypt1111aly ts havt had the opportunity of an ltiring Also at this time I had to provide a practiral f eld cipher for the Indian Army This and its weakness Hre described in NSA Technical Journal Vol XI No J Summer 19flfl parn 4 and 19 In -- ovemher 192r I retired from the British Regular Army to become a 'ar Office civil servant the fir t of two classified as Signal Computers It was arranged that I should tav on in Simla In 1927 I took eight month leave in England durini which I worked for three month at GC CS then in a house in ueen s ate In 1928 the Hussians for the first time introduced one time pads but we were not able to lo very much with them There were two sizes roe cs ICE SSGSSI J t Doc ID 6618711 F XPF RTF NC F 'l a 11 lines of five 5-tigure group i e 275 figures used with messages of not more than 550 figures of intermediate text and b 11 lines of ten 5-figure groups i e 550 figures for longer messages but any message exceeding ll0O had to be divided into parts The operator was permitted to use a pad twice and no more -The pads were used boustrophedon and therefore there were depths of two but reading in opposite directions as the additives were arranged in an odd number of lines We were hardly able to read anything at all except in the case of one or two very stereotyped proforma messages One anecdme I must tell in reference to our earlier more sucre sful days In 1926 I here were two Russian dpher clerks in Kabul named Kolluv and Serafimowi h The all er was I he less n liahle of the t wn and when suffering from frequent morning hangovers made o many mistakes in encipherment that eventually an order came from Moscow that in future all cipher message must be signed in cipher by the clerk responsible rom this time on messages were signed either Zachifrowan Kotlov or Zachifrowan Serafimowich which presented us each time with 40 or i additive figures Hut the day came when Serafimowich himself deciphered a message ordering him to return to Moscow as his papers were not in order He tied at once for sanctuarv to the British Emba y but was ejected and I regret to say was never heard of again In 1929 the War Office decided it needed a military section in GC CS and I as chosen to st art it There had been a naval section there ever since the end of World War I the residue of 40 OB I left India at the end of 1929 having been with Colonel Jeffery for eight and one-half years during which I had learned how to parry his sharp wit or divert it on to other rncl we had become firm friends At the beginning of March 19aO I joined Gr r s again this time as head of the newly formed military section hut m the War Oftice payroll Captain F 4 Jacob afterwards Colonel Jacob who later succeeded me as SlJKLO Washington in 1954 retired from the Army and joined me as deputy He had in 1925 won a prize in the first of two cryptanalytic competitio 1 open to army officer in India which I had instituted and prepared under the auspices of the Intelligence Branch of the General Staff in India I wa3 also allotted an establishment of four posts to be held for three yearn by regular armv officers normally with language qualifications seconded for training in military cryptanalysis Except in the case of Italian which I left entirely to Jacob there was at the time vinually no military intercept to work on and I set mvself the objective of getting as much cryptanalytic training and experience as possible for myself and my officers by collectini cipher systems that others couldn't or wnuldn't deal with This policy was criticized more drnn or1tof' by my paymaster in the War Office but there was really no TOP SFG S IP J H TTLTMA'- a_ltt'rnative if we w1rnted tu gain experiern e in cryptanalysis and particularly erypt analytil' diag11osi Tht dn i ion tu use army officers tu form a reserve in case of war wa not an unqualified suecess Only one of the first batch Pritchard was with me through World War IL Another of them was recalled against my wishes during the Aby sinian War at a critical stage in his career and he suffered professionallv as a result The Government Code and Cipher School had by now moved to Uroadwa Buildinl s part of which it occupied in onjunction with vi 16 until we moved to Bletchley Park in 19 ' 9 A G Denniston was 1till L 1rector Ernst Fetterlein wa still in my opinion far the best general-purpose cr ptanulvst Oliver Struchcy held the title of Chief Cryptographer which he retained until I took it over in 1942 He had hcen in the War Ottice Cork S reet department during World War I and had a considerable reputation for ucce ses both during the war and since its end J E S Cooper who joined together with four or five other young men in 192fi bt'came the tir t head at the Air ction in 19 36 There were also 10 or II survivor from World War I most of them normally engaged in the reconstruction of code book Much of my activities in London from 19 0 until the outbreak of war in September 1939 has already appeared in the NSA Tl'chnirnl d11urrw The s' or_ of our -olution of the various Comintern ciplu r 19 31 thrrnigh J9 t4 q pr1trPd in th ' T - l 11irnl Jo11mr Vol 'Ill - -1 Fall 196 l in an artil'lt' ThP TK'vPlopmPnl of thP A lclitivP nf whi 1 it forms the first part My connection with Japane e cipher and the Japanese langua11 e was described in some detail in the Journal Vol XI - _ Summer 1966 pages 4-9 I took no part in the British olution of the Japanese Hed machine The direct part in this was taken hy Hugh Foss who Joined us about W l2 Foss al o performed the tir t analv i in the British otticc of the German commerc al Oigma machine t I le wa tl1e brother 1 f 1 -ttorn Stra hey the famnu hiof rnphPr 9 JOB SSCREI EXPERIENCES ii IP Doc ID 6618711 TABLE A TABU C A Ofi 122 ' 7 50 'lfi97 P rn JR 4- 111 B 4269 C l 'i 4 ' 4819 i I 17516678 T08 657'i583 Y 14 263J ' 4 67 8 rl X r 04 s 79 u1 'i268 E 18 20 32 46 G o 72 91 2 14 01 19 17 II Ill l 74 Ji II 62 9'299 1 294170 1 0971 102 -101 1 115 73 12147 5 l KJ 1'2 JI l8314Y S Jti H 56 1 0 07 18 2 44 60 819 J o Jl 84 II IO 19 51 X9 2 1 2 3 11 20 2 1 l J 07 II 01 i r o l 11 07 01 18 17 22 19 11 oroaor 19 1 3059 J 'Z7808fi 1113 17 3656 76t4 08 u T BLE D 3 5 l l 14 20 23 S2461 1 19 -h06 nr nun J H TILTMAN 19 II 07 2'2 04 -- 1 Xtrn 18 17 07 11 07 04 IA 17 2'2 11 II 07 11 07 Ill 1907 4 18 Ii 2'2 19 10 19 07 II II I' 0 TABLE E T BLE R 1 2 l 1 4 10 O' 8 H 0 11 12 t J 14 J' @ J 4 Io jo4 j 5 G 04 16 lo 18 19 20 21 2 2 1 24 12 LY 1 14 j 07 9 4 l2 11 07 01 18 17 2'2 19 _ o r o 11 o r 11 07 04 JR 17 11 n U4 I 17 ti 19 JJ Oi 04 16 17 22 19 11 07 04 18 17 22 19 o 0 -19 I' 10 cl II 01 19 17 04 11 l7 17 18 1 I l l '9 11 19 141 CC_ _ _ __ 2 1_ _ _ ____ _ 19_0 cc_ _ c 1s'-- c n_ c 9_ __ TOR 555957 6 l I 21 15 17 16 7 12 24 llJ 8 18 ill ill fi l Ill 6 21 Hi Otl 2 1 9 I I ll Or I 07l04 l l7 1 11 J 2 l II 01 11 07 04 I 17 2 19 22 iO 2 2 II I' o 1 04 lR 6 22 II tr 1 14 1 10 19 11 fil 10 12 13 11 01 -04 18 17 22 19 H '------'l '-0 19 07 11'--'0 1 00 a 29a - 23'- -'l l ----------__ 11 ' D C I l 11 IRR C IF o 11111111ui111111111111111tl lllllllll IIIIIHRIIIIIIIIIIIIHllllllllllllllll IIIIIIIIII E XPERIENCES JOO SFCPEI Iftt 22 2 UNCLASSIFIED Doc ID 6618711 T BLE F 114 5 13 18 3 ll 20 1 6 19 21 l 17 6 7 i2 24 lO ij 8 9 4 2 10 L2 11 13 L' ll lfi 17 1- 19 22 l 20 21 24 A' l DJSTEMP BSNONRI TFEEDS 'f BLE G EXTRACT FRO' VI CONVERSION TABLE 07 06 10 15 24 ll 48 50 6 'l 72 81 8 A 61 ll ' ' O F GA N 27 P l 19 22 40 91 RA l63I I5I07l24l81I08 TABLJ H I6 2 7 J 9 712 8 3 f 1 fi I l O 5 8 2 21 4 9 1 2 5 311 ' 9 4 O 418 6 1 t 2 1 R 2 o s11 2 r 2 7 Io 4 o 8 3 ll o 2 2 0 OI 7 0 9 I 0 8 roe sscnrz anon 12 13 Non - Responsive UNCLASSIFIED This document is from the holdings of The National Security Archive Suite 701 Gelman Library The George Washington University 2130 H Street NW Washington D C 20037 Phone 202 994-7000 Fax 202 994-7005 nsarchiv@gwu edu