(cid:11)(cid:12)(cid:10)(cid:13)(cid:17)(cid:18)(cid:16)(cid:15) Some comments on the Arabic block in Unicode Tom Milo, DecoType Summary 1. Some Extended Arabic characters are typographical variants of characters already adequately covered by the corresponding Basic Arabic Characters; 2. 0626 ARABIC LETTER YEH WITH HAMZA ABOVE is actually a representation form of two nominal characters: 06D5 ARABIC LETTER AE followed by 0621 ARABIC LETTER HAMZA; 3. The graphemes for the aspirated phonemes of Urdu should be added; 06BE ARABIC LETTER HEH DOACHASHMEE can then be deleted or ignored; 4. Urdu Noon-e-ghunna needs fourfold display forms; 5. The characteristic Urdu digits are missing; 6. The Ottoman Kaf-i-Turki (with additional stroke below the main tail) is missing 7. 0644 0644 0647 ARABIC LIGATURE LI-LLAH ISOLATED FORM is missing 1. Doublets: Teh Marbuta, Feh, Qaf, Kaf, Heh, Yeh In some cases it appears to me that Unicode - knowingly - confuses regional calligraphic or typographic variants for encodable characters. E.g., the encoded HEH GOAL, along with its associated GOAL variants, is clearly an attempt to tweak the Naskh typeface (as used for printing the Unicode Standard) to look like Nastaliq: an Eastern variant of the Arabic script. Characters concerned: Basic Arabic Extended Arabic doublets 0629 ARABIC LETTER TEH MARBUTA 06C3 ARABIC LETTER TEH MARBUTA GOAL 0641 ARABIC LETTER FEH 06A2 ARABIC LETTER FEH WITH DOT MOVED BELOW 0642 ARABIC LETTER QAF 06A7 ARABIC LETTER QAF WITH DOT ABOVE 0643 ARABIC LETTER KAF 06A9 ARABIC LETTER KEHEH 0647 ARABIC LETTER HEH 06C1 ARABIC LETTER HEH GOAL 064A ARABIC LETTER YEH 06CC ARABIC LETTER FARSI YEH In fact, this "goal" effect is not at all obligatory for Naskh, as is illustrated here (second row: he-yi- hewwez): 1 Illustration taken from S. K. Gorodnikova and L.B. Kibirkshtis, Uchebnik Jazyka Urdu (Urdu Primer), Moscow 1969., Moscow 1969. Also compare the names of numbers 6 and 8 as written in Nastaliq or printed in Naskh in the illustration accompanying the paragraph on Urdu-Indic numbers. The asterisk following the second middle heh leads to a note pointing out that the alternative form of the letter is used to create aspirated consonants1. The same is true for KAF and so-called KEHEH (06A9): they represent one and the same Arabic letter KAF (0643). As for the name keheh: as far as I know it is in use Urdu to denote the aspirated phoneme /kh/ which still lacks a proper character code in Unicode. I believe that the correct treatment of Urdu HEH and KEHEH is to use the regular Arabic HEH and KAF with a properly designed font, in casu Nastaliq, to render the "user-expected" shape. I observed the same phenomenon in 06CC FARSI YEH to complement 064A ARABIC YEH. Arabic YEH with or without dots in final position is a matter of regional and stylistic preference: e.g., in traditional Egyptian typesetting and calligraphy YEH in final and isolated position never had dots. In Persian this traditional style is still the only one allowed, therefore final and isolated YEH with dots do not occur. Use of Magribi variants of FEH and QAF rules out the use of the corresponding Middle Eastern variants of these letters in the same context. They are not entitled to Unicodes and should be dealt with by font designers. I believe using regional flavours of fonts is acceptable, since the differences in are well known to the users and do not bar him or her from understanding raw text. 2. Heh with Yeh Heh with Yeh in Unicode represents the Arabic letter sequence that is associated with a syntactic construction called izafe or izafet, the Persian equivalent of nominal word composition or linking. It occurs also in Tajiki, Pashtu, Dari, Urdu and Ottoman Turkish. A famous example is the name of the British royal diamond: kuh-i-nur , literally mountain-of- light. The -i- in the example is the connecting element, paraphrased here as -of-. In Arabic script such a linking is optionally expressed by 0650 ARABIC KASRA. The "yeh-above" described in 2 Unicode 06C0 ARABIC LETTER HEH WITH YEH ABOVE corresponds to the Persian hemze-yi- muleyyine: the "relaxed hamza"2. It is the ARABIC LETTER HAMZA used when a word ending in the vowel /e/ (best written with 06D5 ARABIC LETTER AE) is connected to the following word. In contemporary Persian the old pronunciation of hiatus or glottal stop (in Arabic: hamz) between certain vowels has evolved into a glide /y/3. For the same phenomenon Ottoman grammarians uses the Persian term hemze-i-izafet (still with -i- instead of -yi-): the "hamza-of-linking"4. Any noun or adjective can be the first element of izafe. Combining AE and HAMZA in one code is just as erroneous as was the combined LAM-ALEF code point of early Arabic code pages. Furthermore, older and conservative modern spelling use the hemze-yi-muleyyine also on top of YEH. From this it follows that the proper solution would be to consider this floating hamza a separate character, so that users have the freedom to use modern or historical spellings. Needless to say that it would also help Ottoman Turkish data processing. To control the positioning in fonts, designers can still substitute a ligature. All of this also applies to 06C2 ARABIC LETTER HEH GOAL WITH HAMZA above (which happens to be pronounced exactly like modern Persian: /yi/). 3. Aspirated consonants in Urdu and Hindi The real Keheh, whose name is, mistakenly, used in the description of the Persian and Urdu doublets of Kaf, is in fact member of a class of aspirated phonemes that with the present encoding can only be encoded by combining the letter with 06BE ARABIC LETTER HEH DOACHASMEE. I propose to replace this composition method with the proper graphemic encodings, following Hindi practice. Urdu and Hindi are closely related languages, if not one and the same language spoken in different cultures, i.e., Islam and Hinduism. Their phonological systems share the distinctive feature of aspiratedness. An authoritative publication like The Worlds Major Languages5 deals with both languages in one chapter: Hindi-Urdu. It gives the following consonant scheme (asp. stands for aspirated): 3 Both Hindi and Urdu can be traced to Sanskrit, the classic language of Hinduism. The discovery in 18th century of this language and its literature lead to foundation of modern linguistic thinking in Europe. On the one hand the realization of its similarity to other classic languages like Greek and Latin signalled the beginning of Comparative Linguistics and General Linguistics with all its consequential spin-offs like the Neo-grammarians, Structuralism, the Prague School, even Generative Transformationalism and, the latest in comparative linguistics, the Nostratic Theory. On the other hand, Sanskrit was not just a passive object of study, it also actively contributed the discipline of phonetics, a key factor in the emergence of modern linguistic thinking6. Against this backdrop it should come as no surprise that Devanagari script is the most accurate phonetic writing system known in history. The Hindi writing system with Nagari inherits from Sanskrit a very precise orthography with a virtual one-to-one relation between phonemes and graphemes. Consequently it recognizes the aspirated consonant phonemes as independent graphemes7: Various publications treat these aspirated phonemes as independent graphemes also in Urdu. First there is a comprehensive study on Arabic writing systems that notices the same one-to-one relationship between phonemes and graphemes as in Hindi8: The same book continues with this grapheme table: 4 The aspirated consonants are composed by writing the letter of the plain consonant followed by the two-eyed heh9. What makes this table interesting is, that, unlike regular grammars that deal with the script only cursorily and leaving a lot of questions, this book shows how well-adapted the Arabic writing system is for Urdu. But the ultimate argument to treat aspirated Urdu consonants as independent graphemes comes from an authoritative Urdu scholar. Professor Mohammed Zakir, in his Lessons in Urdu Script, also begins with giving the traditional, essentially Arabic-Persian alphabet table: Then he adds the following table recognizing , like Mohammad-Reza Majidi, the graphemic status of the combined letters10: 5 This supports the observation I am making: there should be a series of aspirated consonantal graphemes added to the Basic Arabic block in Unicode. The safest model to follow is the set given by Majidi, as it includes combinations that may be not graphemes representing phonemes, but that nevertheless seem to be recognized as such. From this it also follows, that 06BE ARABIC LETTER HEH DOACHASHMEE can be deleted from Unicode or at least ignored. As a result a much cleaner subset for Urdu can be created without ambiguities that such as between 0647 ARABIC LETTER HEH, 06BE ARABIC LETTER HEH DOACHASHMEE (misplaced representation form) and 06C1 ARABIC LETTER HEH GOAL (doublet), since the latter two can both be deleted or ignored. In order to satisfy user expectation, ARABIC LETTER HEH DOACHASHMEE can still feature on the keyboard layout to allow the user to construct the real Unicodes in a way that may even turn out to be intuitive. Finally. I believe that 06C2 ARABIC LETTER HEH GOAL WITH HAMZA ABOVE is totally redundant, as HAMZA ABOVE is in fact the floating hamza of izafe (para 1), and HEH GOAL is a doublet. For Microsoft, with its already over-crowded Windows Code Page 1256 ARABIC LETTER HEH DOACHASHMEE could be maintained. It would mean that, for the sake of economy, this code page keeps track of the keyed-in sequences, rather than storing the proper grapheme codes. When converting to Unicode the resulting sequences of [plain letter] plus [aspiration mark] must, of course, be replaced by the proper codes. 4. Noon Ghunna also has non-final forms 06BA ARABIC LETTER NOON GHUNNA is given two representation forms: FB9E ARABIC LETTER NOON GHUNNA ISOLATED FORM FB9F ARABIC LETTER NOON GHUNNA FINAL FORM However, it can be documented to have also an initial and a middle form with an optional distinction mark from the regular NOON in these positions. The first illustration shows the four positional variants, the initial and middle of which are identical to those of the regular NOON: 6 Illustration taken from S. K. Gorodnikova and L.B. Kibirkshtis, Uchebnik Jazyka Urdu, Moscow 1969. Translation: Nasal vowels in writing. All nasal vowels are rendered in writing using the letter noon ghunna, which is placed following the character representing the plain short or long vowel or diphthong. This letter (noon ghunna TM) has four written variants, although it can never occur in word-initial position. When connected in initial or middle position a dot is placed above it, as is the case with the letter noon. The second illustration shows the same set with, in non-final positions, an additional distinctive element reminding of a breve mark: Sample taken from Mohammed Zakir, Lessons in Urdu Script, Delhi 1973. Also compare the names of number 5 as written in Nastaliq or printed in Naskh in the illustration accompanying the paragraph on Urdu-Indic numbers. 5. Missing? The Urdu-Indic numbers The series of EXTENDED ARABIC-URDU DIGITS differs from Persian and Arabic. Yet it is nowhere mentioned in the Standard. 7 sample taken from Mohammed Zakir, Lessons in Urdu sample taken from S. K. Gorodnikova and L.B. Script, Delhi 1973. Kibirkshtis, Uchebnik Jazyka Urdu. The digits four, six, seven and nine are markedly different and therefore justify a separate range just like the Persian (Arabic-Indic) digits. On the other hand, I would prefer an approach where all digits should be merged into one series, using regional flavours of fonts instead, since differences in shape of the digits (Arabic, Persian and Urdu) do not bar a user from understanding raw text. 6. The Ottoman Kaf-i-Turki is missing In the late Ottoman era attempts were made to simplify the spelling of Turkish with Arabic letters. From this period stem the Kaf variants with added distinctive features, elaborating the exact graphemic function of the letter concerned. Two of them happen to be covered already in the Unicode, but one seems to have been overlooked, the so-called Turkish kaf, since it seems only to be used there in etymological spellings of the phoneme /y/. Handling Ottoman archives is a serious concern for researchers and governmental institutions alike. Here are some illustrations taken manuals dealing with Ottoman Turkish writing: 8 (from Mahmud Yazir, Eski yazilari okuma anahtari [key to old documents], Istanbul 1942) (from Dr Ali Kemal Belviranli, Osmanlica rehberi 1 [Ottoman handbook], Konya 1996) 9 (from Nahit Tendar & Nebahat Karaorman, Osmanlica okuma anahtari [key for reading Ottoman], Istanbul 1970) 7. Arabic ligature li-llah The glyph ARABIC LIGATURE LI-LLAH ISOLATED FORM and ARABIC LIGATURE LI-LLAH FINAL form [both representing 0644 0644 0647] are missing from the block of Arabic Representation Forms A. This is a major defect: apart from LAM-ALEF , LI-LLAH is arguably the only real ligature of the Arabic script11, since the writing of God's name in Naskh and Naskh-related scrips requires a LAM of reduced height. All Microsoft fonts - with the exception DecoType supplied fonts, that use the Private Area prescribed by the Unicode Standard - apparently assume there must have been some mistake and correct this error partly by replacing the redundant FDF2 [0627 0644 0644 0647] ARABIC LIGATURE ALLAH ISOLATED FORM with FDF2 [0644 0644 0647] ARABIC LIGATURE LI-LLAH ISOLATED FORM. This replacement is justifiable, since the word ALLAH consists of the ligature LI-LLAH preceded by a standard ALEF . The Microsoft workaround may turn out be a rather useful (de facto) standard as it leads to the desired effect in nearly all contexts12. 10