%**start of header \catcode`@=11 % borrow the private macros of PLAIN (with care) \newskip\ttglue \font\elevenrm=cmr10 scaled \magstephalf % roman text \font\eleventt=cmtt10 scaled \magstephalf % typewriter \font\elevenit=cmti10 scaled\magstephalf \font\ninebf=cmbx9 \font\ninerm=cmr9 \font\ninei=cmmi9 \font\ninesy=cmsy9 \font\ninett=cmtt9 \font\ninesl=cmsl9 \font\nineit=cmti9 \font\sixi=cmmi6 \font\sixsy=cmsy6 \font\sixbf=cmbx6 \skewchar\ninei='177 \skewchar\ninesy='60 \hyphenchar\ninett=-1 \def\ninepoint{\def\rm{\fam0\ninerm}% set in nine point families \textfont0=\ninerm \scriptfont0=\sixrm \scriptscriptfont0=\fiverm \textfont1=\ninei \scriptfont1=\sixi \scriptscriptfont1=\fivei \textfont2=\ninesy \scriptfont2=\sixsy \scriptscriptfont2=\fivesy \textfont3=\tenex \scriptfont3=\tenex \scriptscriptfont3=\tenex \textfont\itfam=\nineit \def\it{\fam\itfam\nineit}% \it is family 4 \textfont\slfam=\ninesl \def\sl{\fam\slfam\ninesl}% \sl is family 5 \textfont\bffam=\ninebf \scriptfont\bffam=\sixbf \scriptscriptfont\bffam=\fivebf \def\bf{\fam\bffam\ninebf}% \bf is family 6 \textfont\ttfam=\ninett \def\tt{\fam\ttfam\ninett}% \tt is family 7 \tt \ttglue=.5em plus.25em minus.15em \normalbaselineskip=11pt \setbox\strutbox=\hbox{\vrule height8pt depth3pt width0pt}% strut for 11pt \let\sc=\sevenrm \normalbaselines\rm} \font\scaps=cmcsc10 % for \LaTeX \font\lb=logobf10 \chardef\bs=`\\ % backslash in a string \def\;#1{{\it #1\/}} % The simplest and most useful of all. \input mftmac \parindent=1.5pc % restore after mftmac clobbered it \font\ninetex=cmtex9 \hyphenchar\ninetex=-1 \def\finstring"#1"{\ninetex"#1"\egroup} \def\Beginmft{\par\begingroup\ninepoint % \chardef:=`\: \chardef!=`\! % \chardef==`\= \catcode`\:=12 \catcode`\!=12 \catcode`\==12 \def\mathsemicolon{\mskip\thickmuskip} \let\;=\mathsemicolon % nine-point type: \def\bf{\fam\bffam\ninebf \def\_{\kern.04em\vbox{\hrule width.3em height .6pt}\kern.08em}% \ninebf} \textfont\bffam=\ninebf \scriptfont\bffam=\sixbf \scriptscriptfont\bffam=\fivebf \baselineskip=11pt \def\MF{{\manual hijk}\-{\manual lmnj}} \let\big=\ninebig \setbox\strutbox=\hbox{\vrule height8pt depth3pt width0pt} \rm \setbox\shorthyf=\hbox{-\kern-.05em} \parindent=0pt } \def\Endmft{\par \endgroup} \input tkccode \font\tentk=tkr10 % \font\elvntk=tkr11 \font\tenti=tkti10 % \font\elvnti=tkti11 \def\BeginTurkish{\par \begingroup % \chardef:=`\: \chardef!=`\! % \chardef==`\= \def\`{\char'43}\def\'{\char'100}% \catcode`\:=12 \catcode`\!=12 \catcode`\==12 \sfcode`\:=1000 \sfcode`\!=1000 \sfcode`\==1000 % Frenchspacing after % these } \def\EndTurkish{\par %let's be quite certain about this! \endgroup} \def\verbatim#1{\begingroup\ninepoint \frenchspacing \def\do##1{\catcode`##1=12 } \dospecials \catcode`\:=12 \catcode`\!=12 \catcode`\==12 \parskip 0pt \parindent 0pt \catcode`\ =\active \catcode`\^^M=\active \tt \def\par{\ \endgraf}\overfullrule=0pt \obeylines \obeyspaces \input #1 \endgroup} % a blank line will be typeset at the end of the file; % if you're unlucky it will appear on a page by itself! {\obeyspaces\global\let =\ } \font\titlefont=cmr17 \font\namefont=cmcsc10 scaled \magstep1 \font\twelvebd=cmbx12 \font\sixrm=cmr6 \vsize=43pc \hsize=29pc \parindent=16pt \nopagenumbers \normalbaselines \footline={\hss\sixrm\folio\hss} \def\makefootline{\baselineskip3pc\line{\the\footline}} \def\plainoutput{\shipout\vbox{\makeheadline\pagebody\makefootline}% \advancepageno \vsize=45pc \ifnum\outputpenalty>-\@MM \else\dosupereject\fi} \def\makeheadline{\vbox to\z@{\vskip-22.5\p@ \line{\vbox to8.5\p@{}\ifnum\count0=1 \else\ifodd\count0 \hfill\the\doctitle\ignorespaces \else \the\authorname\hfill\fi\fi}\vss}\nointerlineskip} \newtoks\doctitle \newtoks\authorname \doctitle={The Ottoman Texts Project} \authorname={Walter Andrews and Pierre MacKay} % Added to the general format file % an addition needed by mftmac \def\ninebig#1{{\hbox{$\textfont0=\tenrm\textfont2=\tensy \left#1\vbox to7.25pt{}\right.\n@space$}}} \catcode`@=12 % at signs are no longer letters \tenrm \begingroup \titlefont \the\doctitle\hfil \vskip 3pc \namefont\the\authorname\hfil \tenrm \vskip 1pc {\obeylines Department of Near Eastern Languages and Civilization University of Washington Seattle, Washington 98195 } \vskip 3pc \line{\hfill\twelvebd ABSTRACT\hfill} \vskip 3pc \leftskip 4pc \rightskip 4pc The Turkish orthographic reform of 1928, which required the abandonment of Arabic script in favor of a Latin letter alphabet, was accompanied by a cultural rejection of all literature from the Ottoman period of Turkish history. As a result, only a small part of Ottoman Turkish literature has been made available in scholarly editions in the new orthography. The Ottoman Texts Project is a cooperative effort of Turkish and North American scholars to provide new editions of these works using popular low-priced personal computer systems and standard general purpose software. This paper describes an approach based on the adoption of \TeX\ as the preferred output system for publication. \par \endgroup \vskip 3pc %**end of header \font\eightrm=cmr8 \def\sc{\eightrm} \font\huge=cmr17 scaled \magstep2 \def\dropinitial#1#2{\def\biginitial#1{{\huge#1}}% \def\makeinitial#1{\setbox8\hbox{\strut\vbox to 1.3ex {\hbox{\biginitial#1}\vskip -4pc plus 3.5pc minus 3.5pc}}}% \makeinitial#1% \ifdim\parindent>1.3\wd8\dimen8=\parindent \else\dimen8=1.3\wd8\fi \hangindent=\dimen8\hangafter=-2 \noindent \strut\hskip-1\dimen8\box8{\sc#2}}% \noindent The Ottoman Texts editing and typesetting project represents an attempt to provide a simple, low-cost system for the entry, editing, and typesetting of transcribed [romanized] Ottoman Turkish texts. The purpose of developing such a system was to take advantage of the increasing availability of microcomputers world-wide and to induce the editors of Ottoman texts-- especially Turkish editors--to employ electronic media for their editing tasks. The benefits to scholars of having a large corpus of texts available in machine-readable form seem obvious, but overcoming ``technology cringe'' on the part of scholars whose devotion to medieval literature stems in large part from a strong conservative-traditionalist ideological bent is no small task. Nonetheless, the rewards of converting a significant number of such scholars would be quite high. The vast majority of significant Ottoman Turkish texts await up-to-date editing and the suggested technological change could have a major impact on the speed and accuracy of the editing process as well as on the development of lexicographical tools and on many areas of literary and linguistic study. The situation in Ottoman studies that makes a switch to electronic media especially attractive at this time is rather complex and demands some historical introduction. From its earliest years at about the beginning of the 14th century until early in the 20th century, the Ottoman dialect of Turkish was written in the Arabic script. The political decline of the Ottoman Empire from its pinnacle of world power in the 16th century to its status as a moribund, defeated ally of Germany following World War I, was arrested in the first three decades of this century by a political and ideological revolution that saw the establishment of a Turkish Republic and an accompanying rejection of the literary, cultural, and religious institutions of the Ottoman past. One aspect of the cultural revolution was the adoption of a latin letter alphabet for Turkish, a change which had among its consequences the expansion of literacy beyond a small elite circle to the general populace, a conscious effort to simplify the written language, and a resultant major decline in the ability to read and comprehend the Ottoman literary language in any of its forms. The ethos of the early years of the Republic, to which the Ottoman Empire appeared as decadent and its culture as derivative, also meant that, at a time when the scholarly edition of older texts was becoming a growing concern in other parts of the world, in Turkey interest in things Ottoman, including Ottoman texts, was considered backward, anti-nationalist, counter-revolutionary, and wrong-headed. As a result, very few texts were adequately edited and the population in general was further cut off from its historical past. Since the Second World War, however, there has been an increased scholarly interest in Ottoman texts and in the transcription and edition of such texts. This interest has grown with the growth of a tolerance for some reemergences of older ethical, religious, and cultural practices and attitudes. It is clear that the particular situation in Turkey today lends itself to the adoption of editing methodologies that take advantage of computer technology: there is a large cadre of well-educated persons with very positive attitudes toward technological innovation; the Latin alphabet is used [with modifications for Ottoman transcription]; most of the basic editing work remains to be done; there is already great interest in the types of concordancing, indexing, lexicographical analysis, etc. that can be most easily done by computers. Nonetheless, Ottoman studies is still an area that attracts persons who would be least likely to welcome technological innovation and so any change would need to bring immediate and obvious benefits. When the editing project was initiated, it was decided that the result should have the following characteristics: \medskip \noindent1.\quad It should be easy to use even for the most unsophisticated user. \noindent2.\quad It should be adaptable to many different circumstances and should be easily supportable. \noindent3.\quad It should obviously eliminate the need for more than one entry of the basic text. [This is, of course, common to all computer word-processing systems but it is such a major departure from the usual round of draft typings that its benefits must be emphasized to those who have not experienced it.] \noindent4.\quad It should be capable of producing typeset camera-ready copy for printing. [This is a major potential benefit even in Turkey where the costs of more labor-intensive typesetting methods are growing rapidly.] \medskip The project developed in several stages and was not without its problems and false starts. The first stage involved convincing a noted Turkish scholar and respected editor of Ottoman texts to come to the University of Washington to attempt to edit the collected poems of a 16th century Ottoman poet using the IBM XT already employed by the Department of Near Eastern Languages and Civilizations for the development of Turkish character-sets. Scholarly processes being what they are it turned out to be easier to bring the scholar than to have the necessary word-processing capabilities ready when he arrived. As a result, a rather cumbersome combination of Microsoft's WORD, Rosesoft's ``smart key'' program [PROKEY], and a series of BASIC programs developed by Robert Blum of the UW administration was used to enable the Turkish visitor to input and edit about 90\% of the poems in the collection [over 500 poems] in about three months. The editor, who had had no previous experience of computers and no particular liking or aptitude for them, was an eager and willing convert to the process. Prior to his departure, we were also able to employ a simple translation program which converted the character-set designed for the XT to \TeX\ notation and, subsequently, to produce a typeset sample of the edited text on the SUN minicomputer. The reaction of our visitor to the results of this process, which was carried out with the help of two fellow scholars without the intervention of typists or typesetters, was pure delight and amazement. In the ensuing months the project has been considerably refined and improved. With the invaluable assistance of the UW Humanities and Arts Computing Center and its resident character-sets guru, Gerald Barnett, we have been able to develop a word-processing system that is simple, efficient, flexible, and low-cost. The system is based on Quicksoft's PC WRITE program used with EGA/VGA and compatable graphics hardware.\footnote{$^*$}{NOTE: At present, the system produces a host of irritating ``ghost diacritics'' when used with the IBM PS 2 graphics--these are a distraction more than a real hinderance but, as yet, we have no idea why they occur.} The advantages of PC WRITE for this kind of word-processing are numerous but it is worth mentioning a few in some detail. Given the goal of making this technology widely available among scholars and students [especially among foreign students and scholars], the fact that PC WRITE is low-cost, share-ware [\$89.00 with full support] makes it an attractive alternative. Moreover, PC WRITE permits virtually limitless customization of keyboards, fonts, printer controls, etc. in a manner accessible to persons without any knowledge of programming or programming languages. Using a simple set of programs--a program designed at Duke University for the creation of characters for display on an EGA driven monitor and a program being developed by Gerald Barnett of the UW for the production of downloadable printer fonts--we have been able to produce a word processing system that can display and edit an extended IBM character-set, which will allow the use of modern [roman alphabet] Turkish, the romanized transcription of Ottoman Turkish [Arabic alphabet], and a full English characters font. One can also switch instantly between a standard IBM keyboard, an IBM keyboard adapted to Turkish characters, and the standard Turkish keyboard with extensions for the Ottoman character-set. In addition, the system supports draft printing on the IBM Pro Printer and letter quality printing on the NEC and Toshiba 24 pin printers [with the use of a bi-directional tractor]. The extended IBM character-set uses 8 of the special European characters, 32 special Ottoman transcription characters [on ASCII codes 192--223], and 9 special modern Turkish characters [on ASCII codes 225--233], as well as the full English set. All of the modern Turkish characters appear as characters on the modified standard keyboard. The Ottoman Turkish characters [standard English characters with diacritics] are called up by two-key sequences. For example, a ``d'' with a dot under it is produced by striking ``/'' followed by ``d''; all other special Ottoman characters are produced by the same sequence [``/''$+$``character'']. Keyboard arrangement and the particular character used to call up the special characters can be easily modified to suit the preferences of the user. One fortunate aspect of the print control features of PC WRITE for this project is that the print control program can be set up to support two different fonts for each character. Therefore, \TeX\ notation can be provided as an alternative for each character and translation from the usual word-processor font to \TeX\ notation can be done automatically by simply printing in the \TeX\ input character set to another file. Because PC WRITE produces ``clean'' ASCII files, the material is immediately ready for typesetting in whatever \TeX\ environment is being used. \medskip \centerline{{\elevenrm Accented character sets in \TeX}} \smallskip \noindent In the few years since the official release of \TeX, a number of attempts have been made to adapt the program to languages other than English. The best known successes have depended on adaptations of the program itself, partly because the standard release of \TeX\ can support only one system of hyphenation at a time, which makes a truly bilingual document quite difficult to produce. These adaptations may be broadly classed as program-based extensions of the language. The extension which is most obviously necessary is the addition of a primitive which can control the switch between one predigested hyphenation pattern and another. Michael Ferguson's bilingual CNRS-\TeX, which was initially developed for an environment in the province of Quebec, where French and English are constantly intermingled, is one of the outstanding developments in this class of adaptations, and there are others as well. A second extension is needed to get around the problem of hyphenation in languages which make use of diacriticals and accents. The basic form of \TeX\ will reject any word containing an accent from the evaluation routine which normally looks for acceptable hyphenation breaks. In effect, any word with an accent is treated as if it were an unbreakable horizontal box, and is not evaluated for hyphenation at all. This can make line-breaking very difficult, and several users of \TeX\ have found it necessary to introduce a loop into the program so that accents and diacriticals will be stripped out just before the entry to the hyphenation routine, and then returned to their remembered positions after the discretionary hyphen nodes have been inserted into the word. The disadvantage of both these systems is that the adapted program is no longer \TeX. It is often possible to add the extra features in such a way that the resultant program will produce {\tt DVI} files that are indistinguishable from those generated by \TeX, but the extra features are not generally available on all systems which run \TeX, and the user is often excluded, therefore, from some of the most popular small system versions of \TeX. An alternative solution to the problem of accented languages, though not of bilingual hyphenation patterns, is a font-based, rather than a program-based approach. Font characters may be generated with the accents already applied, and mapped into unused or little-used areas of the normal Computer Modern font table. If these characters are then supplied with an appropriate \TeX\ {\tt\bs lccode} value, the hyphenation loop will recognize them as part of a sequence capable of being hyphenated. For a monolingual application in a language which makes intensive use of accents and diacriticals, this can be an attractive approach, especially when there are reasons for wishing to preserve the ability to make use of small system versions of \TeX. This is the approach we have taken for Turkish \TeX. Turkish provides a delightfully vivid set of examples of accentuation and hyphenation. The Latin-letter character set which has been in use since the orthographic reform of 1928 is extended, even in Modern Turkish, by means of a considerable number of diacriticals and accents. A diligent search through the modern dictionary will produce several five- and six-letter words in which every character is accented, and an intensive search might come up with words as much as nine letters long with every character accented. In critical editions of Ottoman texts, the number of accents more than doubles. Modern Turkish knows only the accented and unaccented pair of letters `{\bf s}' and `{\bf\c s}', but Ottoman Turkish has `{\bf s}', `{\bf\c s}', `{\bf\d s}' and `{\bf\b s}', which represent four completely distinct characters in the Arabic alphabet. The letter `{\bf h}' shows almost as much variety, and so do several others. Our Ottoman Turkish font has twenty-seven accent and letter composites, in addition to the basic twenty-six simple Latin letters. Moreover, all composites can exist in upper case forms as well as in lower case. When a character set is as heavily accented as this, it is desirable to make sure that the accents are positioned over their letters as exactly as possible. The {\tt\bs accent} primitive in \TeX\ does a remarkably good job of positioning accents, but it depends on a very general algorithm, and tends to place accents exactly centered over or under the affected character, no matter what the appearance of that character may be. Donald Knuth recognized this limitation in the very earliest stages of the development of \TeX, and has consistently recommended that frequently used combinations of character and accent be developed as composite single images in the font. The center of a character is not always the best visual position for an accent; top accents should often be slipped just a bit to the right, and bottom accents just a bit to the left of the mechanically defined centerline of the character. Height and depth of accents are similarly subject to aesthetic judgement. The{\tt\bs accent} primitive of \TeX\ works very well indeed for sparsely occurring accentuation, but not so well when accents occur in every second word. The problem of hyphenation in Turkish is even more striking. Turkish is known as an ``agglutinating'' language, which means, in effect that each discrete logico-syntactic qualification of a basic word is expressed in a single syllable tacked onto all the other syllables in the word. At the same time, it is a language in which consonant clusters are virtually unknown. A Turkish word is made up of simple open and closed syllables, of the form {\tt cv} or {\tt cvc}, and in native words there is not even the distinction between long and short vowels. The result is a language in which word-length tends to be greater than it is even in English, and where, as a result, hyphenation is often necessary. The hyphenation rules are inherited from the syllabification of Arabic. A syllable is assumed always to consist of an initial consonant (even when that consonant is no longer written) and to terminate in a vowel or in the next unvowelled consonant. This pattern is followed so absolutely that it is permitted to break up native Turkish suffixes. The plural suffix \;{ -ler-} will be hyphenated as \;{-le-rine} in an environment where the {\tt -cv-cv-cv} pattern predominates. A set of hyphenation patterns for Turkish will therefore be quite simple to produce, but it will have no effect on most Turkish words unless something is done about the problem of accents. A word such as \;{\c cektirilebilecek} ought to provide six discretionary hyphenation nodes: \;{\c cek-ti-ri-le-bi-le-cek}, but the {\tt\bs accent} primitive applied to the first letter will guarantee that the standard version of \TeX\ gives up any attempt to hyphenate it at all.\footnote{$^*$}{The word is a future participle, and describes something as being capable of being extracted at some time in the future--like a tooth. A morphological division of the word would produce a very different hyphenation pattern, \;{\c cek-tir-il-e-bil-ecek}, with only five nodes.} If the initial letter `{\bf \c c}' were a single character in a special font, and were provided with an {\tt\bs lccode} value, the {\tt\bs accent} primitive would no longer appear, and the word could be evaluated for hyphenation. Since the majority of \TeX\ users will never have to deal with {\tt\bs lccode}s at all, a word of explanation is in order here. \TeX\ is designed to take care of the problems of typesetting in a general manner, independent of the language of the text to be set. The program recognizes that while many languages have paired upper and lower case character sets, not all do, and the order of the basic text character set may not be that of the Latin alphabet. For this reason, specific upper and lower case pairings are not built into the program, but are supplied by macro definitions in {\tt plain.tex}. Like all other definitions in plain.tex, they may be replaced, and it is quite possible to dispense with plain.tex altogether, and substitute another basic format file such as {\tt sadece.tex}, {\tt franc.tex}, {\tt einfach.tex} or {\tt sketo.tex}. (Knuth insists, for obvious reasons, that the one thing you may not call it is ``plain.tex.'') If additional characters such as the accented letters of Turkish are made part of the basic input coding table, then they are likely to exist in upper and lower case pairs. Each lower case code is given itself as a lower case {\tt\bs lccode}, and the code of its upper case equivalent as its {\tt\bs uccode}. These can be used to force conversion from one case to the other, but the {\tt\bs lccode} serves an additional purpose. When \TeX\ enters the program loop which searches for discretionary hyphen nodes in each word, it first unpicks all ligatures such as {\bf ffi} and then evaluates the resultant list from the beginning, working on any given word only so long as every character it finds has a valid {\tt\bs lccode}. Any node that is not a simple character with a valid {\tt\bs lccode} causes the routine to terminate; the sequence so marked is supplied with no discretionary hyphen nodes at all, and therefore cannot be broken by the line-breaking algorithm. This is what prevents hyphenation in the case of the Turkish word given above. \medskip \centerline{\elevenrm Input Code Interpretation} \smallskip \noindent The Turkish text-editing system described above is driven from a keyboard mapped to conform as closely as possible to the standard Turkish typewriter keyboard. This mapping is not used directly in the design of the Ottoman Turkish font and, in its present form, is isolated from the actual \TeX\ input. After the raw input has been corrected, it is passed through a filter which converts the accented characters into character pairs (or, in a very few instances, into \TeX\ command sequences). These pairings are based on a proposal made more than ten years ago at the Orientalist Congress held in Paris, in 1974. Owing to the extraordinary richness of the Ottoman Turkish character set, it has been necessary to extend the old proposal, but it still retains the original principles, which are closely associated with the coding scheme used by the Onomasticon Arabicum project. The Onomasticon Arabicum uses a post-positive dot and a post-positive hyphen to indicate diacriticals, which is acceptable in a data-base of names, but not in continuous prose text. To provide the indications for Ottoman Turkish diacriticals, we have taken over the exclamation point `{\tt!}', the equals sign `{\tt=}', and the colon `{\tt:}'. The exclamation point is used for all the ``emphatic'' letters of the Arabic alphabet (the alphabet in which Turkish was written until 1928). These are the letters \;{\d Dad} (usually pronounced as `{\bf z}' in Turkish, and hence paired with a non-Arabic letter known as \;{\.Zad}), \;{\d Sad}, \;{\d Ha'}, \;{\d Ta'} and \;{\d Za'}. The equals sign is used for all the consonants which are represented in Latin-letter transcriptions by a letter with a bar under, such as {\bf\b d} (\;{dhal}), more commonly written in Turkish as `{\bf\b z}', and also for vowels with a macron or, following the Turkish convention, a `hat' accent, and similar forms, chosen like the cupped `{\bf\u g}', because the equals sign is visually closer than the colon is. (Moreover, the colon is needed for a different variety of the letter `{\bf g}'.) The colon is a catch-all for everything else, but works out rather well visually, as it happens. The three post-positives are not accents, but regular characters, which use the \TeX\ convention of ligatures to invoke accented characters from the font, just as the second `{\bf f$\,$}' in the normal \TeX\ `{\bf ff}' ligature pair does. If a standard Latin-letter character does not have an associated ligature table in the font, a following colon will be unaffected. Thus, the letter `{\bf o}', when followed by a colon will produce `{\bf\"o}', but the letter `{\bf e}' when followed by a colon will produce `{\bf e:}'. The equals sign is returned to its normal function in math mode, and the colon and exclamation point can be invoked by the command sequences {\tt\bs:} and {\tt\bs bang} when the simple character will not work. This set of conventions produces an input file which can, if necessary, be edited on a ordinary terminal lacking the special Turkish character features, and which a Turkish speaker can become accustomed to without too much difficulty. When coupled with a well-designed macro file and a rewritten hyphenation table, it provides the possibility of naturalizing a \TeX\ environment into Turkish without any large investment in special purpose hardware and rewritten versions of non-standard (non-)\TeX. \medskip \centerline{\elevenrm The Font} \smallskip \noindent Donald Knuth's Computer Modern fonts come with a wide range of accents, which cover most of the requirements for Turkish. The only obvious lack is the flat cup which is used under both upper and lower case `{\bf h}' as an aesthetic variant for the simple bar under the letter. All the existing accents in Computer Modern are designed for consistency with the stroke-weights and proportions of the underlying alphabetic characters, and it is therefore very desirable to retain the details of this design in any associated font of accented characters. The vertical and horizontal positions may be altered and, for other languages than Turkish, the angle of acute and grave accents over upper case letters, but the basic proportions of each accent or diacritical remain unchanged. This is achieved by taking over the entire text of the Computer Modern character file {\tt accent.mf} and converting the {\tt beginchar $\ldots$ endchar} pairs to {\tt def} and {\tt enddef}. It is not quite so easy as that, but the process is essentially mechanical, and guarantees the preservation of all the essential design details for each accent. (The flat cup under `{\bf h}' is based on the slavic tie accent, turned upside down.) The resultant file, {\tt accdef.mf}, is now full of ``definitions'' which can be invoked as part of the program file for composite characters. Positioning, however, can not be entirely taken care of in the {\tt accdef.mf} file. The accents in {\tt accent.mf} are, for the most part, designed with a fixed reference point at the top of the image, but correct positioning usually requires a knowledge of where the bottom edge will be. It is herefore necessary to take some of the calculations from the accent definitions, and incorporate them into the description of the underlying character. For example, the superscript dot accent in the Computer Modern font is produced as follows. \medskip \Beginmft $\2{iff}\\{ligs}>0\?\2{cmchar}\7"Dot accent";$\par $\2{numeric}\\{dot\_diam}\SH ;\ \\{dot\_diam}\SH =\1{max}(\\{dot\_size}\SH ,% \\{cap\_curve}\SH );$\par $\2{beginchar}(\1{oct}\7"137",5u\SH ,\1{min}(\\{asc\_height}\SH ,\frac10/{7}% \\{x\_height}\SH +.5\\{dot\_diam}\SH ),0);$\par $\2{define\_whole\_blacker\_pixels}(\\{dot\_diam});$\par $\2{italcorr}h\SH \ast \\{slant}+.5\\{dot\_diam}\SH -2u\SH ;$\par $\2{adjust\_fit}(0,0);$\par $\2{pickup}\\{tiny}.\\{nib};\ \\{pos}_{1}(\\{dot\_diam},0);\ \\{pos}_{2}(\\{dot% \_diam},90);$\par $x_{1}=x_{2}=.5w;\ \\{top}\,y_{2r}=h+1;$\par $\2{if}\\{bot}\,y_{2l}<\\{x\_height}+o+\\{slab}\?y_{2l}:=\1{min}(y_{2r}-% \\{eps},\\{x\_height}+o+\\{slab}+.5\\{tiny});\3{fi}$\par $y_{1}=.5[y_{2l},y_{2r}];\ \\{dot}(1,2);\ \9 dot\par $\2{penlabels}(1,2);\3{endchar};$\par \Endmft \medskip \noindent The corresponding {\tt accdef.mf} definition is \medskip \Beginmft $\2{def}\\{dot\_accent}(\2{suffix}\$,@)(\2{expr}\\{dotY\_shift})=$\par $\2{save}@;$\par $\2{forsuffixes}\\{\$\$}=@,@_{\_}\?\2{transform}\\{\$\$};\3{endfor}$\par $\2{numeric}\\{dh}\SH ;\ \\{dh}\SH :=\1{min}(\\{asc\_height}\SH ,\frac10/{7}% \\{x\_height}\SH +.5\\{dot\_diam}\SH );$\par $\2{define\_whole\_blacker\_pixels}(\\{dh},\\{dot\_diam});$\par $\2{pickup}\\{tiny}.\\{nib};\ \\{pos}_{@\_1}(\\{dot\_diam},0);\ \\{pos}_{@\_2}(% \\{dot\_diam},90);$\par $x_{@\_1}=x_{@\_2}=x_{\$};\ \\{top}\,y_{@\_2r}=\\{dh}+1;$\par $\2{if}\\{bot}\,y_{@\_2l}<\\{x\_height}+o+\\{slab}\?y_{@\_2l}:=\1{min}(y_{@% \_2r}-\\{eps},\\{x\_height}+o+\\{slab}+.5\\{tiny});\3{fi}$\par $y_{@\_1}=.5[y_{@\_2l},y_{@\_2r}];$\par $\2{numeric}\\{dot\_span};\ \\{dot\_span}=\\{dh}-\\{bot}\,y_{@\_2l};$\par $@=\\{identity}\2{if}\\{dotY\_shift}<>0\?\6{shifted}(0,\\{dotY\_shift}+\\{dot% \_span})\3{fi};$\par $\2{for}n=1,2\?\2{forsuffixes}e=l,,r\?$\par \quad$z_{@}[n]e=z_{@\_}[n]e\6{transformed}@;\3{endfor}\3{endfor}$\par $\\{dot}(@_{1},@_{2});\ \9 dot\par $\2{penlabels}(@_{1},@_{2});\3{enddef};$\par \Endmft \medskip To get this into position over the letter `{\bf o}', requires the following program text, \medskip \Beginmft $\2{cmchar}\7"The letter dotted o";$\par $\\{dot\_sharp\_values};$\par $\2{beginchar}(\1{oct}\7"025",9u\SH ,\\{dot\_top}\SH ,0);$\par $\2{italcorr}\frac1/{3}[\\{x\_height}\SH ,\\{asc\_height}\SH ]\ast % \\{slant}-.5u\SH \2{if}\\{serifs}\?+.25\\{dot\_diam}\SH \3{fi};$\par $\2{adjust\_fit}(\2{if}\\{monospace}\?.5u\SH ,.5u\SH \2{else}\?0,0\3{fi});$\par $\\{penpos}_{1}(\\{vair},90);\ \\{penpos}_{3}(\\{vair}',-90);$\par $\\{penpos}_{2}(\\{curve},180);\ \\{penpos}_{4}(\\{curve},0);$\par $x_{2r}=\1{hround}\1{max}(.5u,1.25u-.5\\{curve});$\par $x_{4r}=w-x_{2r};\ x_{1}=x_{3}=.5w;\ y_{1r}=\\{x\_height}+\1{vround}1.5\\{oo};\ y_{3r}=-\\{oo};$\par $y_{2}=y_{4}=.5\\{x\_height}-\\{vair\_corr};\ y_{2l}:=y_{4l}:=.52\\{x% \_height};$\par $\2{penstroke}\\{pulled\_arc}_{e}(1,2)\AM \\{pulled\_arc}_{e}(2,3)$\par \quad${}\AM \\{pulled\_arc}_{e}(3,4)\AM \\{pulled\_arc}_{e}(4,1)\AM \1{cycle};\ \9 bowl\par $\2{numeric}\\{dot\_shift},\\{dot\_top};$\par $\2{define\_whole\_blacker\_pixels}(\\{dot\_diam},\\{dot\_top});$\par $\\{dot\_shift}=0;\ \9 in this case, the position happens to be correct\par $x_{7}=x_{1}-.8\\{dot\_diam};\ x_{8}=x_{7}+1.6\\{dot\_diam};$\par $\\{dot\_accent}(7,a,\\{dot\_shift});$\par $\\{dot\_accent}(8,b,\\{dot\_shift});$\par $\2{penlabels}(1,2,3,4,7,8);\3{endchar};$\par \Endmft \medskip \noindent in which the line \medskip \Beginmft $\\{dot\_sharp\_values};$\par \Endmft \medskip \noindent expands to a macro \medskip \Beginmft $\2{def}\\{dot\_sharp\_values}=$\par $\2{numeric}\\{dot\_diam}\SH ;\ \\{dot\_diam}\SH =\1{max}(\\{dot\_size}\SH ,% \\{cap\_curve}\SH );$\par $\2{numeric}\\{dot\_top}\SH ;\ \\{dot\_top}\SH =\1{min}(\\{asc\_height}\SH ,% \frac10/{7}\\{x\_height}\SH +.5\\{dot\_diam}\SH );$\par $\!\3{enddef};$\par \Endmft \medskip \noindent which repeats some of the calculations made in the definition of the dot accent. The composites that result from this programming effort look, for the most part, identical to the results of the application of the {\tt\bs accent} primitive to characters in the regular Computer Modern fonts. The one major difference comes in the shape of the ``hat'' accent over the letter `{\bf i}'. In this instance, the accent would spread beyond the left and right side bearings of the underlying character and mess up the letter spacing if it were not pinched in, so a special narrow hat accent is provided for `{\bf i}'. The proportions of each stroke remain essentially the same as those in the original model, but they form an acute angle over the top of the letter. Except in the case of this character and some of the uniquely Turkish dotted uppercase letters, it will probably be difficult to distinguish the two styles of accent in the final printed version even when they are intermingled in the same text. The creation of the composite characters is only the first stage in the development of the font. Next, the italic correction must be set for all the italic and slant fonts. This is the spacing that may be added to the right side of any slanted character to prevent it from running into something like a non-slanted closing parenthesis. There does not seem to be any way except visual inspection to discover an acceptable italic correction. One wants a fairly simple, general calculation, but one which will do rough justice to all slanted versions of the character. There were more proof copies generated to get the italic correction right than for any other feature of the font. (In the absence of any accessible system on which proofs could be displayed on the screen, a great many paper proofs had to be generated.) Following this comes the generation of ligature and kerning tables, which are necessarily quite large, and need to be carefully worked out since there is only a finite region of a {\tt tfm} file that can be devoted to them. The smaller of the two ligature tables, for the italic fonts, is shown in appendix A. It still needs one further refinement; the kernings appear in the order of English letter frequency, and it might be possible to gain a little efficiency by rearranging some of them. Notice that the `{\bf f$\,$}' ligatures are altogether eliminated. In Turkish it is essential to retain the distinction between the dotted and the undotted `{\bf i}, which cannot be done if the `{\bf fi}' ligature is used. The problem that arises, in fact is to provide adequate separation between the dotted `{\bf i}' and a preceding `{\bf f$\,$}'. In addition to the accented characters, it was necessary to design three additional characters for Ottoman Turkish. The simplest is a dot at about the bar height of lower case `{\bf e}'. This is used for a type of Persian suffix known as ``izafet,'' which is very common in Ottoman texts. The remaining two characters are representations of the Arabic letters ``Ayn'' and ``Hamza,'' which are conventionally represented by opening and closing single quotes in most fonts. The ``lazy man's \;{`ayn}'' (as just illustrated) is acceptable for the occasional reference, but not for extensive literary texts. Ayn is not an accent, it is a regular consonant of the Arabic alphabet and Hamza, though it can be omitted in many positions is also a consonant. What is needed is a pair of characters which are clearly distinguishable from single quotes, but sufficiently like them to conform with the general appearance of Computer Modern. The programs shown below, draw on the same standard definition as is used to generate the single quotes, but alter the position and the proportions. The bulb is uppermost in both instances, and is somewhat smaller than the bulb of the close quote. The tail is brought out further from the side of the bulb, and is tucked more tightly under. The versions for slanted and italic fonts use some special transformations to insure that the {\tenti \`ayn} (that was the character from the Ottoman font) is correctly formed. In effect, the character is built out to the left of the centerline, with a reverse slant, and then reflected back into the normal letter space. The program for these characters is Appendix B. A complete passage from our first proposed critical text edition is given below, first in \TeX\ input coding, and then as typeset. The text from which this passage was extracted runs to twelve pages, and was set without the benefit of a properly rewritten hyphenation table. By good luck, most of the English pattern hyphenations turned out to correspond with acceptable Turkish hyphenations, but it will certainly be necessary to make up a proper Turkish hyphenation table in the near future. When that is done, and an appropriate set of formatting macros has been written to isolate Turkish text from non-Turkish text and math mode, we will have a Turkish language adaptation of \TeX\ which can be exported onto any small \TeX\ system, with no alteration of the program whatsoever. The full range of standard Computer Modern font styles will be available, and will blend in perfectly with the normal unaccented library of Computer Modern fonts. We will not have a truly bilingual version of \TeX, but for a predominantly Turkish language environment we will be offering a cheaper and more accessible monolingual font-based adaptation. \bigskip {\verbatim {Tktexinput} } \bigskip \BeginTurkish \tentk \`A=s:ik!lik! zama=ninda \`is:k! va=sit!asi ve s:eyda=lik! \`a=lemi:nde s:evk! vesi=lesi:, vus!lat eyya=minda mah!abbet muk!tez:a=si, fi:ra=k! gu:nleri:nde h!urk!at i:k!ti:z:a=si, baha=r mev\-si:mi:nde s!oh!bet germi:yyeti:, mah!bu=blar mecli:si:nde s:ara=b keyfi:yyeti:, ca=na=neler i:bra=mi ve \`a=s:ik!lar i:k!da=mi ve fuz:ala= mus!a=h!abeti: ve \`uk!ala= i:lti:fa=ti, ehl..i: di:ller rag:beti: ve t!a=li:bler mi:nneti: i:le di:du:gu: ebya=t ve es:\`a=r, ki: her bi:ri:nu:n= lat!i=f ma\`a=ni=si: ca=m..i naz!ma s:ara=b..i rengi=n ve s:i=ri=n h=aya=la=ti bezm..i: s!afa=da nuk!l..i: s:ekkeri=n olup mu\`a=s:i:ra=n..i mecli:s..i: z=evk! bu meyh=a=nenu:n= ba=deci:si: ve h!ari=fa=n..i bezm..i: s:evk! bu ka=s:a=nenu:n= sebu=-kes:i: olmis:lardi. K!alem..i: i:\`ti:z=a=r bu h=a=me..i: i:nki:sa=r i:le bu evra=k!a tah!ri=re i:k!da=m ve bu ecza=ya tast!i=re i:hti:ma=m go:sterdi:. \EndTurkish \tenrm \vfil\eject \strut\vskip .75in \centerline{{\elevenrm Appendix A}} \smallskip \centerline{{\elevenrm The {\eleventt turkit.mf} driver file}} \bigskip \Beginmft \input turkit.tex \Endmft \vfil\eject \strut\vskip .75in \centerline{{\elevenrm Appendix B}} \smallskip \centerline{{\elevenrm The {\tt aynhmz.mf} file}} \Beginmft \input aynhmz.tex \vfil \Endmft \bye