1

ProposalforaMalayalamScriptRootZoneLabelGenerationRuleset(LGR)LGRVersion:3.0Date:2018-09-25Documentversion:1.7Authors:Neo-BrahmiGenerationPanel[NBGP]

1. GeneralInformationThepurposeofthisdocumentistogiveanoverviewoftheproposedMalayalamLGRinthe XML format and the rationale behind the design decisions taken. It includes adiscussionofrelevantfeaturesofthescript,thecommunitiesorlanguagesusingit,theprocess and methodology used, the repertoire of code points included, variant codepoint(s),wholelabelevaluationrulesandinformationonthecontributors.Theformalspecification of the LGR can be found in the accompanyingXMLdocument: proposal-malayalam-lgr-25sep18-en.xml. Labels for testing can be found in the accompanyingtextdocument:test-labels-malayalam-25sep18-en.txt

2. ScriptforWhichtheLGRIsProposedISO15924Code:MlymISO15924KeyN°:347ISO15924EnglishName:MalayalamLatintransliterationofnativescriptname:malayāḷaṁNativenameofthescript:മലയാളംMaximalStartingRepertoire(MSR)version:MSR-3

3. BackgroundonScriptandPrincipalLanguagesUsingItMalayalamisaDravidianlanguagewithabout38millionspeakersspokenmainlyinthesouthwestof India,particularly inKerala, theLakshadweepIslandsandneighbouringstates,andalsoinBahrain,Fiji,Israel,Malaysia,Qatar,Singapore,UAEandtheUK.

Malayalam was first written with the Vatteluttu alphabet (വെ)ഴു,് Vaṭṭeḻuttŭ),whichmeans'roundwriting'anddevelopedfromtheBrahmiscript.TheoldestknownwrittentextinMalayalamisknownastheVazhappalliorVazhappallyinscription,isintheVatteluttualphabetanddatesfromabout830AD.

AversionoftheGranthaalphabetoriginallyusedintheCholakingdomwasbroughttothe southwest of India in the 8th or 9th century and was adapted to write theMalayalam and Tulu languages. By the early 13th century it is thought that a

2

systematized Malayalam alphabet had emerged. Some changes were made to thealphabet over the following centuries, and by the middle of the 19th century theMalayalamalphabethadattaineditscurrentform.

AsaresultofthedifficultiesofprintingMalayalam,asimplifiedorreformedversionofthe script was introduced during the 1970s and 1980s. The main change involvedwritingconsonantsanddiacriticsseparatelyratherthanascomplexcharacters.Thesechanges are not applied consistently so the modern script is often a mixture oftraditionalandsimplifiedletters.

Thescripthasthefollowingnotablefeatures:

● Malayalam script is written left to right in horizontal lines using a syllabicalphabet inwhichall consonantshavean inherentvowel.Diacritics,which canappear above, below, before or after a consonant, are used to change theinherentvowel.

● When they appear at the beginning of a syllable, vowels are written asindependentletters.

● Chillaksharam is another feature of Malayalam. A chillu is a pure consonantwithouttheuseofavirama,whichkillstheinherentvowelofaconsonant.

● When certain consonants occur together, special conjunct symbols are usedwhichcombinetheessentialpartsofeachletter.

3.1 TheEvolutionofMalayalamScriptMalayalam was first written in the Vatteluttu alphabet, an ancient script of Tamil.However,themodernMalayalamscriptevolvedfromtheGranthaalphabet,whichwasoriginallyusedtowriteSanskrit.BothVatteluttuandGranthaevolvedfromtheBrahmiscript,butindependently.

3.2 VatteluttualphabetVatteluttu (Malayalam:വെ)ഴു,്, Vaṭṭeḻuttŭ, “roundwriting”) is a script that hadevolved from Tamil-Brahmi and was once used extensively in the southern part ofpresent-dayTamilNaduandinKerala.

Malayalam was first written in Vatteluttu. The Vazhappally inscription issued byRajashekharaVarman is theearliest example,dating fromabout830CE. In theTamilcountry,themodernTamilscripthadsupplantedVatteluttubythe15thcentury,butintheMalabar region,Vatteluttu remained ingeneraluseup to the17th century,or the18thcentury.Avariant formof thisscript,Kolezhuthu,wasuseduntilabout the19thcentury mainly in the Kochi area and in the Malabar area. Another variant form,Malayanma,wasusedinthesouthofThiruvananthapuram.

3

3.3 Grantha,TigalariandMalayalamscriptsAccordingtoArthurCokeBurnell,oneformoftheGranthaalphabet,originallyusedinthe Chola dynasty, was imported into the southwest coast of India in the 8th or 9thcentury, which was then modified in course of time in this secluded area, wherecommunicationwiththeeastcoastwasverylimited. It laterevolved intotheTigalari-Malayalam script used by theMalayali, HavyakaBrahmins andTuluBrahmin people,but was originally only applied to write Sanskrit. This script split into two scripts:TigalariandMalayalam.WhileMalayalamscriptwasextendedandmodifiedtowritethevernacularMalayalam language, Tigalari was used for Sanskrit only. InMalabar, thiswritingsystemwastermedArya-eluttu(ആര0എഴു,്,Āryaeḻuttŭ),meaning“Aryawriting”.(SanskritisanIndo-AryanlanguagewhileMalayalamisaDravidianlanguage).

Vatteluttuwasingeneraluse,butwasnotsuitableforliteratureinwhichmanySanskritwordswereused.LikeTamil-Brahmi,itwasoriginallyusedtowriteTamil,andassuch,didnothavelettersforthevoicedoraspiratedconsonantsusedinSanskritbutnotusedinTamil.Forthisreason,VatteluttuandtheGranthaalphabetweresometimesmixed,asin theManipravalam literature (a literary style used in medieval liturgical texts in South India). One of the oldest examples of this, Vaishikatantram (ൈവശികത78ം,Vaiśikatantram), dates back to the 12th century, where the earliest form of theMalayalamscriptwasused,but itseemstohavebeensystematizedtosomeextentbythefirsthalfofthe13thcentury.

ThunchaththuEzhuthachan,apoet fromaroundthe17thcentury,usedArya-eluttutowrite his Malayalam poems based on Classical Sanskrit literature. For a few lettersmissing in Arya-eluttu (ḷa, ḻa, ṟa), he used Vatteluttu. His works becameunprecedentedlypopulartothepointthattheMalayalipeopleeventuallystartedtocallhimthefatheroftheMalayalamlanguage,whichalsopopularizedArya-eluttuasascripttowriteMalayalam.However,Granthadidnothavedistinctionsbetweeneandē,andbetween o and ō, as itwas only used towrite the Sanskrit language. TheMalayalamscript as it is today was modified in themiddle of the 19th century when HermannGundertinventedthenewvowelsignstodistinguishthem.

Bythe19thcentury,oldscriptslikeKolezhuthuhadbeensupplantedbyArya-eluttu–that is the currentMalayalam script. Nowadays, it is widely used in the press of theMalayalipopulationinKerala.

Malayalam andTigalari are sister scripts descended from theGrantha alphabet. Bothsharesimilarglyphicandorthographiccharacteristics.

3.4 OrthographyreformIn1971,theGovernmentofKeralareformedtheorthographyofMalayalambypassingagovernmentordertotheeducationdepartment.Theobjectivewastosimplifytheuseofprint and typewriting technology of that time, by reducing the number of glyphs

4

required.In1967,thegovernmentappointedacommitteeheadedbySooranadKunjanPillai the editor of the Malayalam Lexicon project. It reduced the number of glyphsrequired for Malayalam printing from around 1000 to around 250. The abovecommittee's recommendations were furthermodified by another committee in 1969[105].

Noneof themajornewspapers implemented it completely.Buteverynewspaper tookitsownsubsetfromtheproposal.Thereformedscriptcameintoeffecton15April1971(theKeralaNewYear),byagovernmentorderreleasedon23March1971.

3.5 LanguagesusingtheMalayalamscriptThescriptisalsousedtowriteseveralotherlanguagessuchasPaniya,BettaKurumba,andRavula (all atEGIDS5).TheMalayalam language itselfwashistoricallywritten inseveraldifferentscripts.

NBGPconsideredlanguageswithEGIDSscale1to4forinclusion.MalayalamisoneofthetwolanguageswritteninMalayalamscript(vizMalayalam&Sanskrit)meetingthiscriterion. Malayalam is placed among the 22 scheduled languages of India. Sanskrit,although it falls under EGIDS 4, is not considered in Malayalam script LGR becauseMalayalamisrarelyusedtowriteSanskrit.

3.6 ZWJ/ZWNJApart fromtheexistingUnicodecharactercodepoints inMalayalam[110],ZeroWidthJoiner(ZWJ,U+200D)andZeroWidthNon-Joiner(ZWNJ,U+200C)arewidelyusedtocontrol how ligatures are formed. Being invisible characters, they are often removedwhiledoingnormalization,particularlybeforedoingastringcomparison,orcollation.ICANN'sMaximal StartingRepertoire (MSR) for IDNLGR is based on these exclusionrulesforZWJandZWNJ.[101]

Impact of excluding them fromdomainname system:Although IDNA2008 allowstheuseofZWJandZWNJindomainnames,theyarenotallowedintherootzonelabels,duetoexclusionfromMSR.

Hence it isnotpossible toregisterMalayalamdomainnameswithwordsthatcontainzwj/zwnj.

Therearethreecases:

● MissingZWNJ isconsideredasaspellingmistake.Example:TamilNadu(tamiɭnadu)iswrittenas: തമി9നാ;[0D240D2E0D3F0D340D4D200C0D280D3E0D1F0D4D](correct),[0D240D2E0D3F0D340D4D0D280D3E0D1F0D4D](incorrect).But there are no identified cases where a missing ZWNJ forms another validwordwithdifferentmeaning.

5

● MissingZWJmeans,thewordisadifferentwordwithdifferentmeaning.Thisisvery rare – v aNyavanika (meaning: large curtain)വന0വനിക vanyaVanika (meaning: wild garden) pair is often cited as anexampleforthis.Butmanypeoplearguethisisnotavalidcase.[102][103]

● MissingZWJnevermeansaspellingmistake,but justawritingstyle.Therearemanyexamplesforthis.-ന<(meaning:goodness)isoneobviousone.

Historically,ZWJwasusedtorenderchillu incertain fontsbut laterUnicode includedchillu charactersasstandalonecodepointsandMSR-3also includes these standalonechillucharacters.

Pre-Unicode 5.0, Chillu letters were encoded as a sequence using Joiners. The olderencodingisstillprevalentindata,suchascorporaandmayevenbeincurrentuse.

ButthislegacyrepresentationofChilluusingViramaandZWJisruledoutbecausetherootdoesnotallowjoiners,sothere isno issuewiththeduplicateencodingofChillu.Hence,itistobenotedthatalthoughatomicencodingofChillulettersisnotuniversallyused,RootZoneonlyallowstheatomicencoding.

Figure1:AtomicEncodingMalayalamChillus[107]

ZWNJ,isusedtopreventtheformationofconjunctligaturesanditisrequiredtoavoidspellingmistakesandunnecessary conjuncts.Forexample, ina2-word label, the firstwordendinginviramacanformconjunctwiththesecondwordstartinginaconsonant.Thiscausesaspellingmistake.

3.7 TheStructureofMalayalamScriptTheMalayalamAksharamorgraphemeclusterisbasedontheMalayalamphonologicalsystem,withthefollowingbasicphonologicaltemplate.PhonologyVowels:Malayalamhasfiveshortandfivelongvowels.Vowelsoccurinallpositionsinaword,exceptforo,whichisnotpermittedattheendofit.Italsohastwodiphthongs,ai,au.

6

Figure2:MalayalamVowelPhonology[109]

Consonants:BesidesaDravidianconsonantalinventory,MalayalamhasaspiratedstopsandsupplementarysibilantsborrowedfromIndo-Aryan.[f]occursmostlyinEuropeanborrowings.Voicelessunaspiratedstops,nasalsand laterals [l], [ɭ]canbegerminated.The distinction between single and geminated consonants is phonemic. Only sixconsonants,[m],[n],[ɳ],[r],[l],and[ɭ],canoccurwordfinally.

Figure3:MalayalamConsonantPhonology[109]

Sandhi: internal and external sandhi are commonplace. They result in vowel andconsonantdeletion,assimilationofconsonantsandfusion.Stress:itfallsalwaysonthefirstsyllableofawordScriptandOrthographyMalayalam is written in an abugida script derived ultimately from Brāhmī in whichevery consonant carriesan inherenta.Thealphabeticorder isbasedonphonologicalprinciples: itbeginswiththesimplevowelsanddiphthongs followedby25stopsandnasalsarrangedinfivegroupsaccordingtotheirplaceofarticulation.Itcontinueswithsemivowels (liquids and glides) and fricatives to end in two retroflex liquids whichdon'texistinSanskritand,thus,werenotrepresentedinBrāhmī.Geminated consonants and other consonant clusters are written side by side or oneabovetheother.BeloweachMalayalamsignappearsthestandardtransliterationintheLatinalphabet,andbetweensquarebracketsitsequivalentintheInternationalPhoneticAlphabet.

7

The following sections provide details of the Malayalam sounds and how these arewritteninMalayalam.

ṛisasyllabicvowelfoundonlyinSanskritloanwords.[f]isfoundmostlyinUrduandEnglishloanwordsanddoesn'thaveaspecificsign;itisrepresentedwithphthatalsoservesfor[pʰ].Vowels

Vowelsarewritteninthisformwhentheyareindependentlyused.

അ U+0D05A

ആ U+0D06AA

ഇ U+0D07I

ഈ U+0D08II

ഉ U+0D09U

ഊ U+0D0AUU

ഋ U+0D0BR

എ U+0D0E

ഏ U+0D0F

ഐ U+0D10

ഒ U+0D12

ഓ U+0D13

ഔ U+0D14

8

E EE AI O OO AU

Table1:MalayalamVowelsVoweldiacritics

Vowels can also be written as diacritics referred to as Matras, when these followconsonants. Their forms are given below, illustrated with the letterക (U+0D15)

MALAYALAMLETTERKA.

ക U+0D15KA

കാ U+0D15U+0D3EKAA

കി U+0D15U+0D3FKI

കീ U+0D15U+0D40KII

കു U+0D15U+0D41KU

കൂ U+0D15U+0D42KUU

കൃ U+0D15U+0D43KR

െക U+0D15U+0D46KE

േക U+0D15U+0D47KEE

ൈക U+0D15U+0D48KAI

െകാ U+0D15U+0D4AKO

േകാ U+0D15U+0D4BKOO

കൗ U+0D15U+0D57KAU

Table2:MalayalamVowelDiacriticsConsonants

Malayalamhas the following consonants, generallyarrangedmymannerandplaceofarticulation.

ക U+0D15KA

ഖ U+0D16KHA

ഗ U+0D17GA

ഘ U+0D18GHA

ങ U+0D19NGA

ച U+0D1ACA

ഛ U+0D1BCHA

ജ U+0D1CJA

ഝ U+0D1DJHA

ഞ U+0D1ENYA

ട U+0D1FTTA

ഠ U+0D20TTHA

ഡ U+0D21DDA

ഢ U+0D22DDHA

ണ U+0D23NNA

ത U+0D24TA

ഥ U+0D25THA

ദ U+0D26DA

ധ U+0D27DHA

ന U+0D28NA

പ U+0D2APA

ഫ U+0D2BPHA

ബ U+0D2CBA

ഭ U+0D2DBHA

മ U+0D2EMA

9

യ U+0D2FYA

ര U+0D30RA

റ U+0D31RRA

ല U+0D32LA

ള U+0D33LLA

ഴ U+0D34LLLA

വ U+0D35VA

ശ U+0D36SHA

ഷ U+0D37SSA

സ U+0D38SA

ഹ U+0D39HA

Table3:MalayalamConsonantsAnusvaramandVisargam

Anusvaram: An anusvaram (അനുസlാരം anusvāram), or an anusvara, originallydenoted the nasalization where the preceding vowel was changed into a nasalizedvowel,andhenceistraditionallytreatedasakindofvowelsign.InMalayalam,anusvararepresentedas◌ം(0D02)however,simplyrepresentsaconsonant/m/afteravowel,thoughthis/m/maybeassimilatedtoanothernasalconsonant.Itisaspecialconsonantletter, different from a "normal" consonant letter, in that it is never followed by aninherent vowel or another vowel. In general, an anusvara at the end of aword in anIndian language is transliteratedasṁin ISO15919,butaMalayalamanusvaraat theendofawordistransliteratedasmwithoutadot.

Visargam:Avisargam(വിസർഗം,visargam),orvisarga,representsaconsonant/h/afteravowel,andistransliteratedasḥ.Liketheanusvara,itisaspecialsymbol,andisneverfollowedbyaninherentvoweloranothervowel.InMalayalam,◌ഃ(0D03)isthevisargasymbol.

Chilluletters(Chillaksharam)andSamvruthokarams

IntheIndo-EuropeanfamilyoflanguageslikeSanskrit,alargenumberofwordsendinconsonants. But in Dravidian languages likeMalayalam themajority ofwords end invowels. But, the chillaksharams of Malayalam are exceptions to this general feature.Chillaksharamsarepureconsonants,withoutanyvowelsound.[111]

Chillaksharam is an original feature of Malayalam used only with 6 consonants atpresent.Theconsonantsareന(na),ണ(ṇa),ര(ra),ല(la)ള(ḷa)andക(ka)andtheircorrespondingchillusare ൻ (ṉ), ൺ (ṇ), ർ (r), ൽ (l) ൾ (ḷ) and ൿ (ḳ) incertaincontexts,occurringattheendofthewordwithouttheimplicitvowel.TheChillu0D7Feventhough israre, isstill inuse predominantly inreligiousliteratureand inprepernouns suchasnamesandplacenames.Hence it is included in theLGRto treatChillucharactersconsistently.

10

ൺ U+0D7ANN

ൻ U+0D7BN

ർ U+0D7CRR

ൽ U+0D7DL

ൾ U+0D7ELL

ൿ U+0D7FK

Table4:MalayalamChilluletters

Samvruthokaram is a soft ending virama (chandrakkala). Any consonant can befollowedbyconsonant+◌ു(0D41)+◌◌്(0D4D),creatingthesamvruthokaramformofthatconsonant.InsouthernKerala,theUmatra◌ു(0D41)andchandrakkala(virama)◌◌് (0D4D) together form the grapheme for samvruthokaram. However, in northernKerala, just chandrakkala (visible virama) standing alone is used. The chandrakkalaaloneattheendofawordistreatedasSamvruthokaram.

Chandrakkala coming within a word (followed by other character(s) of the word)denotes a conjunct letter formed by the character(s) preceding and following thechandrakkala.

ExamplesofSamvruthokaram:

(ethumeaningwhich),codepoints-U+0D0FU+0D24U+0D41U+0D4D

(athumeaningthat)codepoints-U+0D05U+0D24U+0D41U+0D4D

For thewords that end in chillu, Samvruthokaram isused tomake thepronunciationclearer.Eithersamvruthokaramisaddeddirectlytotheword-endingchillaksharam,ortheword-endingchillaksharamisgeminatedandSamvruthokaramisaddedtoit.

Thefollowingarethemainphonologicaltransformationsofchillaksharam.[113]

1. The word-ending consonant written as chillaksharam, is geminated and asamvrukthokaramisattached:

2. To the word-ending consonant written as chillaksharam, a samvrukthokaram isattached:

11

3.Thechillaksharamundergoesthesamephonologicalchanges(inprogressive/

regressiveassimilation,gemination,etc.)asinthecaseofotherconsonantsinthe

contextofcombinationofsyllables:

4.Insandhi,whenavowelfollowsachillaksharam,theyjoininthesamewayaswhen

vowelsfollowotherconsonants:

EventhoughSamvruthokarammaybeseenasderivedfromthevowelsഅ(a)orഉ(u),infact,ithasanindependentidentityasavowel.ThisfeatureisseenonlyinMalayalam.[111]Aselectionofconjunctconsonants

AconsonantcanbecombinedwithanotherconsonantorconjunctusingVirama.Conjunctswithmorethanfourconsonantsarerare.Theconjunctu7v0isformedbyfiveconsonants.

kka ṅka ṅṅa cca ñca ñña ṭṭa ṇṭa ṇṇa tta nta nna

NLF wക xക xങ yച

zച

zഞ

;ട {ട {ണ

|ത }ത }ന

LF ~ � � � � � ) � � , 8 �

Table5:MalayalamConjunctConsonants

NLF-Non-ligatedformhasavisiblevirama(chandrakkala)

12

LF-Ligatedforminwhichconsonantsareconjoinedfullyorpartially(asrenderedbyfonts)Conjunctswithdiacriticsusingയ(U+0D2F),ര(U+0D30),ല(U+0D32),വ(U+0D35)

Conjunctconsonantsformedwithയ(0D2F),ര(0D30),ല(0D32)andവ(0D35)arerendered with diacritic marks/signs in the glyph. Examples of these in combinationwithക(0D15)andപ(0D2A)aregivenbelow.Otherconsonantscanbecombinedinsimilarfashion.

Consonant+യ Consonant+ര Consonant+ല Consonant+വ

ക0(0D150D4D0D2F)

7ക(0D150D4D0D30)

�(0D150D4D0D32)

കl(0D150D4D0D35)

പ0(0D2A0D4D0D2F)

7പ(0D2A0D4D0D30)

�(0D2A0D4D0D32)

പl(0D2A0D4D0D35)

Table6:MalayalamConjunctswithdiacriticsusingയ(U+0D2F),ര(U+0D30),ല(U+0D32),വ(U+0D35)

4. OverallDevelopmentProcessandMethodologyThe Neo-Brahmi Generation Panel (NBGP) has been formed from members havingexperience in linguistics and computational linguistics. Under the Neo-BrahmiGenerationPanel, thereareninescriptsbelongingtoseparateUnicodeblocks.Eachofthese scripts is assigned a separate LGR; however Neo-Brahmi GP ensures that thefundamental philosophy behind building those LGRs are all in sync with all otherBrahmi-derivedscripts.

4.1GuidingPrinciples

TheNBGPadoptsthefollowingbroadprinciplesfortheselectionofcode-pointsinthecode-pointrepertoireacrosstheboardforallthescriptswithinitsambit.

4.1.1Inclusionprinciples:

4.1.1.1Modernusage:Every character proposed should be in the everyday usage of a particular linguisticcommunity.CharacterswhichhavebeenencodedinUnicodefortranscriptionpurposesonly or for archival purposes will not be considered for inclusion in the code-pointrepertoire.

13

4.1.1.2Unambiguoususe:Every character proposed should have unambiguous understanding among thelinguisticcommunityaboutitsusageinthelanguage.

4.1.2Exclusionprinciples:

The main exclusion principle is that of External Limits on Scope. These compriseprotocols or standardswhich are prerequisites to the Label Generation Rulesets. Allfurtherprinciplesareinfactsubsumedundertheselimitationsbuthavebeenspeltoutseparatelyforthesakeofclarity.

4.1.2.1ExternalLimitsonScope:The code point repertoire for root zone being a very special case, at the top of theprotocolhierarchies,therangeofavailablecharactersforselectionasapartoftheRootZonecodepointrepertoireisalreadyconstrainedbyvariousprotocollayersbeneathit.Thefollowingthreemainprotocols/standardsactassuccessivefilters:i.TheUnicodeChart:Outofallthecharactersthatareneededbythegivenscript,ifthecharacterinquestionisnotencodedinUnicode,itcannotbeincorporatedinthecodepointrepertoire.Suchcasesarequiterare,giventheelaborateandexhaustivecharacterinclusioneffortsmadebytheUnicodeConsortium.ii.IDNAProtocol:Unicode being the character encoding standard for providing the maximum possiblerepresentation of a given script/language, it has encoded as far as possible all thepossible characters needed by the script. However, the Domain name being aspecialized case, it is governed by an additional protocol known as IDNA(Internationalized Domain Names in Applications). The IDNA protocol introducesexclusionofsomecharactersoutofUnicoderepertoire frombeingpartof thedomainnames.iii.MaximalStartingRepertoire:TheRoot-zoneLGRbeingarepertoireofthecharacterswhicharegoingtobeusedforcreation of the root zone TLDs, which in turn are an even more specialized case ofdomain names, the ROOT LGR procedure introduces additional exclusions on IDNAallowedsetofcharacters.Example:MALAYALAM SIGN AVAGRAHA "ഽ " (U+ 0D3D) even if allowed by IDNA

protocol,isnotpermittedintheRootZoneRepertoireasperthe[MSR].Tosumup,therestrictionsstartoffbyadmittingonlysuchcharactersasarepartofthecode-block of the given script/language. This is further narrowed down by the IDNA

14

2008ProtocolandfinallyanadditionalfilterintheformofMaximalStartingRepertoirerestrictsthecharactersetassociatedwiththegivenlanguageevenmore.

4.1.2.2NoRareandObsoleteCharacters:There are characterswhich have been added toUnicode to accommodate rare formslikeMALAYALAMLETTERVOCALICL"ഌ"(U+0D0C),whichisanobsoletevowelusedto write Sanskrit words and is not considered as part of the modern Malayalamorthography. All such characterswill not be included. This is in consonancewith theConservatismprincipleaslaiddownintheRootZoneLGRprocedure.

5. RepertoireBasedontheLGRProcedurefortheRootZoneandtheMSR,NBGPconductedthecodepointanalysisoftheMalayalamscript.Theanalysisispresentedinthissection,includingthelistofcodepointsrecommendedforinclusionandexclusionfromtherepertoire.

15

5.1 MalayalamsectionofMaximalStartingRepertoire[MSR]Version3

Figure4:MalayalamCodePagefrom[MSR]

Color convention1: Allcharactersthatareincludedinthe[MSR]-YellowbackgroundPVALIDinIDNA2008butexcludedfromthe[MSR]-PinkishbackgroundNotPVALIDinIDNA2008-Whitebackground

5.2 UnicodeCodePointsInclusionThefollowingcodepointsareincludedintherepertoire.

Sr.No.

UnicodeCodePoint

Glyph CharacterName IndicSyllabicCategory

Refs.

1This document needs to be printed in color for this to be read correctly.

16

1 0D02 ◌ം MALAYALAMSIGNANUSVARA Anusvaram [106]

2 0D03 ◌ഃ MALAYALAMSIGNVISARGA Visargam [106]

3 0D05 അ MALAYALAMLETTERA Vowel [106]

4 0D06 ആ MALAYALAMLETTERAA Vowel [106]

5 0D07 ഇ MALAYALAMLETTERI Vowel [106]

6 0D08 ഈ MALAYALAMLETTERII Vowel [106]

7 0D09 ഉ MALAYALAMLETTERU Vowel [106]

8 0D0A ഊ MALAYALAMLETTERUU Vowel [106]

9 0D0B ഋ MALAYALAMLETTERVOCALICR Vowel [106]

10 0D0E എ MALAYALAMLETTERE Vowel [106]

11 0D0F ഏ MALAYALAMLETTEREE Vowel [106]

12 0D10 ഐ MALAYALAMLETTERAI Vowel [106]

13 0D12 ഒ MALAYALAMLETTERO Vowel [106]

14 0D13 ഓ MALAYALAMLETTEROO Vowel [106]

15 0D14 ഔ MALAYALAMLETTERAU Vowel [106]

16 0D15 ക MALAYALAMLETTERKA Consonant [106]

17 0D16 ഖ MALAYALAMLETTERKHA Consonant [106]

18 0D17 ഗ MALAYALAMLETTERGA Consonant [106]

19 0D18 ഘ MALAYALAMLETTERGHA Consonant [106]

20 0D19 ങ MALAYALAMLETTERNGA Consonant [106]

21 0D1A ച MALAYALAMLETTERCA Consonant [106]

22 0D1B ഛ MALAYALAMLETTERCHA Consonant [106]

23 0D1C ജ MALAYALAMLETTERJA Consonant [106]

24 0D1D ഝ MALAYALAMLETTERJHA Consonant [106]

25 0D1E ഞ MALAYALAMLETTERNYA Consonant [106]

17

26 0D1F ട MALAYALAMLETTERTTA Consonant [106]

27 0D20 ഠ MALAYALAMLETTERTTHA Consonant [106]

28 0D21 ഡ MALAYALAMLETTERDDA Consonant [106]

29 0D22 ഢ MALAYALAMLETTERDDHA Consonant [106]

30 0D23 ണ MALAYALAMLETTERNNA Consonant [106]

31 0D24 ത MALAYALAMLETTERTA Consonant [106]

32 0D25 ഥ MALAYALAMLETTERTHA Consonant [106]

33 0D26 ദ MALAYALAMLETTERDA Consonant [106]

34 0D27 ധ MALAYALAMLETTERDHA Consonant [106]

35 0D28 ന MALAYALAMLETTERNA Consonant [106]

36 0D2A പ MALAYALAMLETTERPA Consonant [106]

37 0D2B ഫ MALAYALAMLETTERPHA Consonant [106]

38 0D2C ബ MALAYALAMLETTERBA Consonant [106]

39 0D2D ഭ MALAYALAMLETTERBHA Consonant [106]

40 0D2E മ MALAYALAMLETTERMA Consonant [106]

41 0D2F യ MALAYALAMLETTERYA Consonant [106]

42 0D30 ര MALAYALAMLETTERRA Consonant [106]

43 0D31 റ MALAYALAMLETTERRRA Consonant [106]

44 0D32 ല MALAYALAMLETTERLA Consonant [106]

45 0D33 ള MALAYALAMLETTERLLA Consonant [106]

46 0D34 ഴ MALAYALAMLETTERLLLA Consonant [106]

47 0D35 വ MALAYALAMLETTERVA Consonant [106]

48 0D36 ശ MALAYALAMLETTERSHA Consonant [106]

49 0D37 ഷ MALAYALAMLETTERSSA Consonant [106]

50 0D38 സ MALAYALAMLETTERSA Consonant [106]

18

51 0D39 ഹ MALAYALAMLETTERHA Consonant [106]

52 0D3E ◌ാ MALAYALAMVOWELSIGNAA Matra [106]

53 0D3F ◌ി MALAYALAMVOWELSIGNI Matra [106]

54 0D40 ◌ീ MALAYALAMVOWELSIGNII Matra [106]

55 0D41 ◌ു MALAYALAMVOWELSIGNU Matra [106]

56 0D42 ◌ൂ MALAYALAMVOWELSIGNUU Matra [106]

57 0D43 ◌ൃ MALAYALAMVOWELSIGNVOCALICR

Matra [106]

58 0D46 െ◌ MALAYALAMVOWELSIGNE Matra [106]

59 0D47 േ◌ MALAYALAMVOWELSIGNEE Matra [106]

60 0D48 ൈ◌ MALAYALAMVOWELSIGNAI Matra [106]

61 0D4A െ◌ാ MALAYALAMVOWELSIGNO Matra [106]

62 0D4B േ◌ാ MALAYALAMVOWELSIGNOO Matra [106]

63 0D4D ◌് MALAYALAMSIGNVIRAMA Chandrakkala/Virama

[106]

64 0D57 ◌ൗ MALAYALAMAULENGTHMARK Matra [106]

65 0D7A ൺ MALAYALAMLETTERCHILLUNN ChilluLetters [106]

66 0D7B ൻ MALAYALAMLETTERCHILLUN ChilluLetters [106]

67 0D7C ർ MALAYALAMLETTERCHILLURR ChilluLetters [106]

68 0D7D ൽ MALAYALAMLETTERCHILLUL ChilluLetters [106]

69 0D7E ൾ MALAYALAMLETTERCHILLULL ChilluLetters [106]

70. 0D7F ൿ MALAYALAMLETTERCHILLUK ChilluLetters [106]

Table7:MalayalamCodePointRepertoire

5.3 CodePointSequenceThe following sequenceshavebeendefined for thepurposeof variantdefinitionsand

WLErules(seesection6.1).

19

1. U+0D28U+0D4DU+0D31 ന ◌് റ [}റ]

MALAYALAMLETTERNAMALAYALAMSIGNVIRAMAMALAYALAMLETTERRRA

2 U+0D33U+0D4DU+0D33 ള ◌് ള[�]

MALAYALAMLETTERLLAMALAYALAMSIGNVIRAMAMALAYALAMLETTERLLA

3 U+0D7BU+0D31 ൻ റ [ൻറ]

MALAYALAMLETTERCHILLUNMALAYALAMLETTERRRA

Table7a:MalayalamCodePointSequences

5.4 UnicodeCodePointExclusionThefollowingcodepointsareexcludedbecausetheyarearchaicorobsoleteincurrentMalayalamorthography.

Sr.No.

UnicodeCodePoint

Glyph CharacterName IndicSyllabicCategory

Reason

1. 0D0C ഌ MALAYALAMLETTERVOCALICL

Vowel ഌ(0D0C)anobsoletevowelusedtowriteSanskritwords.Theletterഌisveryrare,andarenotconsideredaspartofthemodernMalayalamorthography.

2. 0D44 ◌ൄ MALAYALAMVOWELSIGNVOCALICRR

Matra ◌ൄ(0D44)isthematrasignofobsoletevowelVOCALICRRൠ(0D60)whichisnotamongtheapprovedcodepointsinMSR-3.ItisnolongerusedinMalayalamorthography.

3. 0D29 ഩ MALAYALAMLETTERNNNA

Consonant ഩ(0D29)correspondstoTamilṉaன.Usedrarelyinscholarlytextsto

20

representthealveolarnasal,asopposedtothedentalnasal.[108].Inordinarytextsbotharerepresentedbynaന(0D28).

Table8:MalayalamExcludedCodePoint

6. VariantsThis section discusses the variant code points found in Malayalamwithin script andwithotherrelatedscripts.

6.1 In-scriptvariantsThissectionlistssequencesthatshouldbeconsideredvariantsofoneanother.Set# Characters CodePoints Glyph

1. a)} + റ 0D28+0D4D+0D31 or

b)ൻ + ◌് + റ 0D7B+0D4D+0D31

c) ൻ + റ 0D7B+0D31 ൻറ

2. a) �+ള 0D33+0D4D+0D33 �

b) ള+ള 0D33+0D33 ളളTable9:In-scriptVariantAnalysis

Set1:Thesearevariouswaystowritetheconjunct“nta”inMalayalam.1a)Herentais

encodedasacombinationof0D28+0D4D+0D31anditisrenderedas in most of

theMalayalamUnicodefontsandafewoftheMicrosoftfontsrenderitas}റ.

1b)ishowsomeMicrosoftfontshaveencodednta0D7B+0D4D+0D31anditis

renderedas inthosefontsandas inotherfonts.Becauseofthe

renderingproblem,itissafetodisallowthissequencebyWLE.(PleaseseeWLERule1).

Although1.c)hasalsobeenusedhistoricallytowritentaandsuchsequentialstyleof

21

writingisstillinuse,thatcombinationcanalsobeusedtowritenrainwordslike

(Henry)or(Enrica).[112]Hencethesequenceof1.c)isallowed.Thevariants

inset1containonlytwosequences:0D7B+0D31and0D28+0D4D+0D31.Andthe

dispositionis“blocked”.

Set 2: The consonantള (0D33) rarely follows anotherള inMalayalam, except in thecase of some place names. The double conjunct ofള (0D33) formed by code points0D33+0D4D+0D33isrenderedastheglyph�whichlooksvisuallyverysimilartoaളfollowinganotherള.Thiscanresultinspoofedlabels.Forexample,inMalayalamwewrite“vellam”as“െവ�ം”-0D350D460D330D4D0D330D02(meaning:water),aspoofedlabelcanwriteitas“െവളളം”-0D350D460D330D330D02.Thisshouldbeblocked.

However,thispatterngivesrisetosomecomplicationsbecauseiteffectivelymakesthe

Halant (0D4D)avariantof a "nullposition", in this case,whenever itoccursbetween

twoinstancesof0D33ളLLA.Variantdefinitionsofthatnaturecanleadtounexpected

resultsbecausealabel:0D330D4D0D330D4D0D33canbeanalyzedtwoways:

{0D330D4D0D33}{0D4D}{0D33}and{0D33}{0D4D}{0D330D4D0D33}

NBGPtakesintoaccountthedataprovidedbytheIPonoccurrenceofthesesequences

incertainlabelswhereaconsonantള(0D33)followsanotherള:IPhadfoundthatthe

frequency issmall.However, thecommunity feedbackshowsan increase inusagedue

to foreign-language-borrowedwords language. The detailed analysis and supporting

datacanbefoundinAppendixC.

Therefore,NBGPhasdecidednottodefineSet2asvariants,buttohandlethiscaseby

usingaWLErule.The rulewillnotallowa first consonantള (0D33) ina label tobe

immediatelyfollowedbyasecond0D33,butrequiresanH(0D4D)(oranyothereligible

codepoint)tointervene.

Asequence0D330D4D0D33hasbeendefinedintherepertoiresection.Addingsuchasequencewouldhavetheeffectofallowingthecase“�ള”(0D330D4D0D330D33)while continuing to disallow0D330D33 everywhere else, including in “ള�” (0D330D330D4D0D33).

22

6.2 Cross-ScriptVariantsTheMalayalamcharactersintablesbelowareconsideredvariantcodepointswithsomecharactersinOriyaandTamilastheycouldbeconsideredvisuallysamefortheusers.SeeAppendixA foradditional codepoints forother scriptswhicharevisually similarbutnotconsideredasvariantcodepointsforthereasonslisted.

6.2.1 Cross-scriptvariantsforTamilandMalayalam

VariantSet Tamil Malayalam

CP Glyph CP Glyph

1. 0B9C ஜ 0D1C ജ

2. 0BB5 வ 0D16 ഖ

3. 0BAE ம 0D25 ഥ

4. 0BBF ◌ி 0D3F ◌ി

5. 0BC6 ெ◌ 0D46 െ◌

6. 0BC7 ே◌◌் 0D47 േ◌

Table10:Tamil–MalayalamCrossScriptVariants

6.2.2 Cross-scriptvariantsforOriyaandMalayalamCaseofMalayalamandOdia(Oriya)TTHAConsonant:

Thisisthecaseof"ConsonantTtha"whichhappenedtoretainthesameshapedespite

beingpartofdifferentscripts,i.e.,MalayalamandOdia.Thesecharactersare:

ഠ-MALAYALAMLETTERTTHA(U+0D20)

ଠ-ORIYALETTERTTHA(U+0B20)

Both characters look exactly alike and belong to a "Consonant" category. As they are

consonants,eachofthem,eveninthesimplestformi.e.thecharactersthemselves,are

validlabels.AspertheNBGPcross-scriptvariantinclusionpolicy(AppendixD),thisisa

validcaseforinclusion.Also,eveniftheyaresinglecharacters,whenthesamecharacter

23

combines,theoreticallytheycanformaninfinite2numberofcross-scriptvariantlabels

betweenthescriptsinvolved.Herearesamplesofsomeofthoselabels:

Malayalam Oriya

ഠഠഠU+0D20U+0D20U+0D20

ଠଠଠU+0B20U+0B20U+0B20

ഠഠഠഠU+0D20U+0D20U+0D20U+0D20

ଠଠଠଠU+0B20U+0B20U+0B20U+0B20

ഠഠഠഠഠU+0D20U+0D20U+0D20U+0D20U+0D20

ଠଠଠଠଠU+0B20U+0B20U+0B20U+0B20U+0B20

Since, having such labels is a realistic possibility and the corresponding labels lookalmostexactlyalike,NBGPhasproposedthem(togetherwithsimilarcombiningmarks)asblockedvariants.

VariantSet Oriya Malayalam

CP Glyph CP Glyph

1. 0B20 ଠ 0D20 ഠTable11:Oriya–MalayalamCrossScriptVariants

7. WholeLabelEvaluation(WLE)RulesThissectionprovidestheWLErulesthatarerequiredbyallthelanguagesmentionedinSection4whenwritteninMalayalamScript.TheruleshavebeendraftedinsuchawaythattheycanbeeasilytranslatedintotheLGRspecifications.BelowarethesymbolsusedintheWLErules,foreachofthe"IndicSyllabicCategory"asmentionedinthetableprovidedforcodepointrepertoireinSection5.

2Though theoretically infinite, this number would be limited to the number of such labels whose equivalent punycode string would not exceed 63 characters including the ACE prefix "xn--".

24

7.1.1 Variablesordefinitions

V → VowelM → Matra(VowelSign)C → ConsonantL → ChilluH → Chandrakkala/Halant/Virama(◌◌്U+0D4D)B → Anusvaram(◌ംU+0D02)X → Visargam(◌ഃU+0D03)

7.1.2 RulesforFormingAksharamRule1:HmustbeprecededbyCortheM◌ു(0D41)(Samvruthokaram)

Rule2:MmustbeprecededbyC

Rule3:BmustbeprecededbyC,VorM

Rule4:XmustbeprecededbyC,VorMRule5:LcannotbeprecededbyB,XorH

Rule6:LabeldoesnotbeginwithL

Rule7:Theള (0D33)cannotimmediatelyfollow ള (0D33)

8. ContributorsNeo-BrahmiGenerationPanel(NBGP)VeenaSolomon(veena.ycet@gmail.com)PrasadPattarumadomKesavaKurup(pkpdelhi@gmail.com)SanthoshThottingal(santhosh.thottingal@gmail.com)AnivarAravind(anivar@indicproject.org)JijoPappachan(jijospeaks@yahoo.com)

9. References[MSR]IntegrationPanel,"MaximalStartingRepertoire—MSR-3Overviewand

Rationale",28March2018https://www.icann.org/en/system/files/files/msr-3-overview-28mar18-en.pdf

[EGIDS]ExpandedGradedIntergenerationalDisruptionScale,

https://www.ethnologue.com/about/language-status(Accessedon5thJuly,2018)

[101] Unicode®StandardAnnex#31MarkDavis,“UnicodeIdentifierAndPatternSyntax”:2.3LayoutandFormatControlCharactershttp://unicode.org/reports/tr31/#Layout_and_Format_Control_Characters(Accessedon5thJuly,2018)

[102] “ReportonMalayalamUnicodeIssues”(2012)preparedbySanthoshThottingal(alsopartofNEGP)andsubmittedtoUnicodeviaWikimediaFoundation.Itdiscussesbothchilluandntaissues:

25

http://thottingal.in/documents/ReportonMalayalamUnicodeIssues.pdf (Accessedon5thJuly,2018)

[103]ഓളംDictionary,https://olam.in/ (Accessedon5thJuly,2018) [104] RoozbehPournaderandCibuJohny,“OldandNewChillusinMalayalamand

implicationsforSinhala” http://www.unicode.org/L2/L2013/13036-chillus-uptake.pdf (Accessedon5thJuly,2018)

[105] Wikipedia,“Malayalamscript” https://en.wikipedia.org/wiki/Malayalam_script (Accessedon5thJuly,2018)

[106] Omniglot,“Malayalam(മലയാളം)” https://www.omniglot.com/writing/malayalam.htm (Accessedon5thJuly,2018)

[107] TheUnicodeStandard,Version10.0.,Chapter12“SouthandCentralAsiaI:OfficialScriptsofIndia”, https://www.unicode.org/versions/Unicode10.0.0/ch12.pdf#page=65 (Accessedon5thJuly,2018)

[108] Everson,Michael(2007)."ProposaltoaddtwocharactersforMalayalamtotheBMPoftheUCS"(PDF).ISO/IECJTC1/SC2/WG2N3494.Retrieved2009-09-09: http://std.dkuug.dk/jtc1/sc2/wg2/docs/n3494.pdf (Accessedon5thJuly,2018)

[109] AlejandroGutmanandBeatrizAvanzati“Malayalam,TheLanguageGulper” http://www.languagesgulper.com/eng/Malayalam.html (Accessedon5thJuly,2018)

[110] MalayalamRange:0D00–0D7F,TheUnicodeStandard,Version11.0https://unicode.org/charts/PDF/U0D00.pdf (Accessedon5thJuly,2018)

[111] R.Chitrajakumar,N.GangadharanRachanaAksharaVedi“SamvruthokaramandChandrakkala”https://www.unicode.org/L2/L2005/05213-samvruktokaram.pdf(Accessedon2ndAugust,2018)

[112] SanthoshThottingal,“}റ-ഭാഷ,യുണിേ~ാ�,ചി7തീകരണം”https://blog.smc.org.in/nta-rendering-rules/(AccessedonAug2nd,2018]

[113] R.Chitrajakumar,N.GangadharanRachanaAksharaVedi“Chillaksharamof MalayalamLanguage”https://unicode.org/L2/L2005/05214-chillu.pdf

(Accessedon27thAugust2018)

26

10. AppendixA:ExcludedIn-ScriptVariantsAsthefollowingformationsarenotvalidasperAksharamformationrules,thesecasesarenotproposedasvariants.

1.ഈ 0D08 ഈ

ഇ+◌ൗ 0D07+0D57 ഇ◌ൗ

2.ഊ 0D0A ഊ

ഉ+◌ൗ 0D09+0D57 ഉ◌ൗ

3.ഔ 0D14 ഔ

ഒ+◌ൗ 0D12+0D57 ഒ◌ൗ

4.ഓ 0D13 ഓ

ഒ+◌ാ 0D12+0D3E ഒ◌ാ

5.ഐ 0D10 ഐ

എ+െ◌ 0D0E+0D46 എെ◌TableA-1:ExcludedIn-ScriptVariantsDuetoInvalidCombination

InTableA-2,Column1:Thesevowelsignshaveglyphpieceswhichstandonbothsidesoftheconsonant;theyfollowtheconsonantinlogicalorder,andshouldbehandledasaunitformostprocessing.Column2:Although,Unicodedefinesthiscanonicaldecomposition,theStandardrecommendsnottousethesequence[107],p501.Therefore,itisnotadvisabletousetheminIDNlabels;theyareblockedherebyaksharaformationrule.

CodePoint1+Glyph1 CodePoint2+Glyph2

െ◌ാ(0D4A) െ◌(0D46)+◌ാ(0D3E)

േ◌ാ(0D4B) േ◌(0D47)+◌ാ(0D3E)

◌ൗ(0D57) െ◌(0D46)+◌ൗ(0D57)TableA-2:SplitVowelCase

27

11. AppendixB:ConfusableCodePointsThecode-pointsbelowarevisuallyconfusingonlyinsmallerfontsandcanbeexcludedfromconsiderationasvariantcodepoints.

Tamil Malayalam

ஸ(0BB8) സ(0D38)

TableB-1:Tamil-MalayalamConfusableCodePoints

Oriya Malayalam

ଂ (0B02) ം (0D02)

ଃ (0B03) ഃ (0D03) TableB-2:Oriya-MalayalamConfusableCodePoints

AttheSriLankaface-to-facemeeting,itwasdecidedtoexcludethecodepointsbelowfromthevariantlistasthesedonotlookalike,duetoround/squarestructuraldifferences.

Kannada Malayalam

ಲ(0CB2) ല(0D32)

TableB-3:Kannada-MalayalamConfusableCodePoints

Telugu Malayalam

ల(0C32) ല(0D32)

TableB-4:Telugu-MalayalamConfusableCodePoints

CodepointsinTableB-5,B-6,andB-7wouldqualifyascross-scriptcodepointvariantsbuttherearenotenoughofthemtoformavariantlabels,thereforethesecasescanbeexcluded. (If only combining marks are variants for a given script, no label can beformedwithoutusingat leastonenon-variant codepoint). In the caseof Sinhala, therelevantbasecharacterisdistinct.

Kannada Malayalam

◌ಂ (0C82) ം (0D02)

◌ಃ (0C83) ഃ (0D03)

TableB-5:Kannada-MalayalamTooFewIdenticalCodePoints

28

Telugu Malayalam

◌ం (0C02) ◌ം(0D02)

◌ః (0C03) ◌ഃ(0D03)

TableB-6:Telugu-MalayalamTooFewIdenticalCodePoints

Sinhala Malayalam

ം (0D82) ം (0D02)

ඃ (0D83) ഃ (0D03) TableB-7:Sinhala-MalayalamTooFewIdenticalCodePoints

NBGP also considers that 0D1F (ട) MALAYALAM LETTER TTA is similar to 0073 (s) LATIN SMALL LETTER S and 0455 (ѕ) CYRILLIC SMALL LETTER DZE. However, Latin script and Cyrillic script are not derived from the Brahmi script. This case is out of scope of NBGP cross script variant analysis.

12. AppendixC:Caseofള(0D33)+ള(0D33)The consonant ള (0D33) rarely follows another ള in Malayalam, except in the case of some place names. The double conjunct of ള (0D33) formed by code points 0D33 + 0D4D + 0D33 is rendered as the glyph � which looks visually very similar to a ള following another ള. This can result in spoofed labels. For example, in Malayalam we write “vellam” as “െവ�ം” - 0D35 0D46 0D33 0D4D 0D33 0D02 (meaning: water), a spoofed label can write it as “െവളളം” - 0D35 0D46 0D33 0D33 0D02.

Combination Codepoints Glyph

�+ള 0D33+0D4D+0D33 �

ള+ള 0D33+0D33 ളള

TableC-1:Caseofള(0D33)+ള(0D33)

This has been restricted by a WLE rule 7. It allows the combination “�ള” (0D33 0D4D 0D33 0D33) which is present in words like “ഉ�ള�” (meaning: inner dimension viz.

29

volume), and blocks the combination “ള�” (0D33 0D33 0D4D 0D33) which is rarely found in usage. The existence of “ളള” (0D33 0D33 ) in considerable percentage on the web can be attributed to misspelling due to extreme visual similarity.

===================================================================

ProposedrecommendationfromtheIntegrationPanel

===================================================================

ProposedrecommendationforMalayalamDATE:2018-06-12

Overview

TheIPrecentlydiscoveredatechnicalissuewiththeproposedvariantsforMalayalam.

IssueStatement

TheMalayalamLGRdefinesthefollowingvariant

0D330D33<->0D330D4D0D33(i.e.:ളള<-->�)

This pattern gives rise to some complicationsbecause it effectivelymakes theHalant(0D4D) a variant of a "null position", in this case, whenever it occurs between twoinstances of 0D33ള LLA. Variant definitions of that nature can lead to unexpectedresultsbecausealabel0D330D4D0D330D4D0D33canbeanalyzedtwoways:

{0D330D4D0D33}{0D4D}{0D33}and

{0D33}{0D4D}{0D330D4D0D33}

Asaresultofthis,variantdefinitionsofthisnature,althoughseeminglywell-definedonthecodepointlevelcanleadtounexpectedvariantrelationsamonglabels.

Therefore, such kinds of variant sequence definitions cannot be used without somefurtherrestriction.BelowtheIPwillsuggesttwopossibleapproachesandrequeststhattheGPconsidertheminlightoftheknowledgeofhowthescriptisused.

Background:

LookingattheMalayalamsamplefiletheIPnotes:0D330D33ളള existsonce(1)insampleof60Klabels(it'spartofthelongerpattern:0D330D4D0D330D33or�ള)

0D330D330D33(ളളള) exists(0)times

0D330D4D0D33(�) exists523times,or.9%ofthetotal;ofthese:

● 1/10or52arefollowedbyan0D4D(Halant):0D330D4D0D330D4D(�്)

30

● none(0)isofthepattern0D330D4D0D330D4D0D33(orlonger)

Fromthisonecanconclude:

● � isquitefrequentandcanbespoofedbyളള (whichdoesn'toccurnormally

oratleastnotfrequently)

● �് alsooccurswith some frequency and could be spoofed byള� (the latter

againnotseeninthesample)

● �ള doesoccur,ifrarely,andcanbespoofedbyള� orളളള,butnotby�്ള(wherethecodepointsare:0D330D4D0D330D33,0D330D330D4D0D33and0D330D4D0D330D4D0D33)

UnderthedefinitionintheproposedLGR�ള andള� arenotactuallyvariantlabelsofeachother,while�്ള isavariantof�ള eventhoughitshouldn'tbe.(Thereasonwhy the last label shouldn'tbeavariant label isbecause the secondhalantwouldberenderedvisibly,makingitdistinct.)

Longer patterns are either rare or do not occur in standard sample; they seemquitelikelytobenonsensical(atleastsomeofthem).Therefore,thecasesseensofarwouldappeartobethetotalsetofcaseswherethereisapracticalneedforsomevariantsorotherrestriction.

Options

The IP identified two suggested options to resolve the issue.

Option One

Restrictingthevariantsoitcannotoccurfollowingan0D33ളorHalant.

Ifthevariantcanbelimitedtothebeginningofacluster,thatis,arequirementaddedthatitonlyapplieswhennotfollowingan0D33of0D4D,thenwecantakestillcareofthe most frequent and second most frequent case, and these cases produce variantlabels thatarerelated inexpectedways: longerstringsofalternating0D33and0D4Dposenoproblemsasanyalternategroupingofcodepointsintosequencesdoesnotleadtoanyadditionalvariants.Onlytheleading{0D330D33}or{0D330D4D0D33}wouldcausevariants.Inparticular�്ള (withavisibleHalant)wouldnotbecomeavariantof

�ള, etc. However, cases like�ള / ള� /ളളള would still not fully work as

intendedasthefirstandsecondlabelwouldnotbevariantsofeachother,andonlythefirstwouldbeavariantofthelast.

OptionTwo

Restrictingvalidlabelstoexcludeളള

31

Restricting labels fromcontaining two0D33ള that arenot joinedbyaHalantwouldrobustly prevent any spoofing. However, it would also disallow a small number ofpotentiallymeaningfullabels.(About0.0015%ofthewordsinthetestfileareaffected-or1in60K).Novariantdefinitionwouldbeneeded.

Recommendation

TheIPrequeststheNBGPtostudytheseoptionsandtoconsiderthemindeterminingaproposedapproach to fixing the issuewith thekindofvariantmappingmentionedattheheadofthedocument.

Werealizethattheserepresentatrade-off.FortheRootZonewefeelcomfortablethatrestrictionoftheallowedlabelstoavoidsomeproblemcasesisdefinitelyappropriate,even if the process contains a String Review phase that would allow the manualweedingoutofspecificbadcases.

However,wefeelthatanoptionthatleavessome,ifrare,opportunitiesforspoofingmaywellbe inappropriate for thesecondandother levels aswell: for those levels,humanoversightoftheprocessisgoingtobeevenlessavailable.

TheIPsuggeststhattheGPalsoweightheextenttowhichdecisionsfortheRootZoneaffectotherzones(byexample).

===================================================================

Feedbackfromcommunity

===================================================================

നീള&മുടി,neelalla mudiis how people say നീളമു� മുടി,neelamullamudi[meaning: long hair, lit. hair with length], locally in Valluvanad area of NorthKerala.Similarly,

നല� താള� പാ)്,nalla thaalalla paattu, is the same as നല� താളമു�പാ)്,nallathaalamullapaattu[meaning:(a)songwithgoodrhythm]

െവ�� കിണ�,vellalla kinaru,is െവ�മു�കിണ�,vellamullakinaru[meaning: (a) well with water]This label is not blocked because�ള isallowed.

I don't think these need to be considered, as theള&partin these labels is aspokencontractionofഉ&,ulla[meaning:having,with].

In other parts of Kerala, the spoken dialect changes the contraction to "െളാ�" orേളാ�whichareallowedaspertherule.

32

Then there are some place names likeമാള&.On doing a Google search, I got onlyasingleresult[google.co.in].

Feedbackfromthecommunity:

Iwon'trecommendaddingsuchrulesbasedontheexistenceofcurrent(andpopular)vocabulary of 2018.Malayalamhas an active practice of borrowingwords fromotherlanguages(mainlyfromEnglishnowadays)ratherthaninventingnativewords.Becauseofthisanythingthatisavalidconjunctcancomeintothelanguage.Hereisanexample:Youmayknow,Iamatypefacedesignertoo.WhensomeofourinitialfontsdidnothavetheOpenType rules to handle�+ബ ,�+ബു, itwas because nobody could find awordthatcanhavesuchacombination.Later,around2010,Facebookbecameathing.People started writing it inMalayalam. Our fonts could not handle the renderinggracefully and then we added the required ligatures and rules and released a newversion. While I was working on another typeface, another conjunct�+മ was notsupportedonthethinkingthatthereisnoMalayalamwordwith�മ.Butlaterafriendcameandcomplainedhewantstohaveanerror-freerenderingforഅ�മീർ..Sothatisaboutthe'reasoningofrareoccurrenceinMalayalam'.Btw,therearepeopleandplaceswith nameമാള&(Malalla) - try a google search.We people from Valluvanad areaoftenhasthisനല�നീള&മുടി,നല/താള&പാ2്,െവ&&കിണ7...

Agooglesearchforെവ&&showsmethatitisaplacenameinIdukki.

Aboutthevisualsimilarity,again,asatypedesigner,weconsciouslymakethemvisuallydifferentwhiledesigning.�+ള->�appearveryjoinedwiththetailsfusedtogether,Whileളളappearwithenoughspacingbetweenthelettersandnofusingoftails.

Also,ററisasimilarcasewherepeoplewritetwoRatogethertoget/tta/,Almostallfontsnowadaysstackthemif it is for/tta/.Butnotguaranteed.Sosimilarargumentscanbethereforthataswell.

Misspellinglikeമീററ7,ലാററൈററ7etc.comestomymind.

Inallthesecases,exclusionruleswouldbetheleastpreferredchoice.

33

13. AppendixD:NBGPCross-scriptVariantInclusionPolicyIf, inanytwogivenscripts,allthepotentialcross-scriptvariantsconsistofdependent(e.g. Vowel Signs, Anusvara, Visarga, Chandrabindu etc.) charactersONLY, then thatentiresetcanbeignoredandnocross-scriptvariantsbeproposedbetweenthosetwoscripts.

If, in any two given scripts, there isAT LEAST ONEnon-dependent (e.g. Consonant,Vowel etc.) cross-script variant character/sequence present, all the potential cross-script variants be considered and proposed between the two scripts.This cross-script analysishasbeen restricted to the scripts thathavedescended fromthe Brahmi as most of them share similar usage patterns. By and large, all of thesescriptshaveacommonsetofcharactersthatexistedinBrahmiscriptandbearthesameidentities.However,asthescriptsbranchedoutfromtheBrahmi,dependingonvariousfactors,theshapesofthecharacterschanged.Thischangeintheshapewasnotuniformacross all the characters and the scripts. Some characters shapes did changesignificantlywhereassomeof themstillretainedsimilarity.Thecross-scriptsimilarityanalysisalsoaimstoidentifysuchcaseswherethesamecharacterretainedalmostthesame shape despite being part of the different scripts. These set of characters arevariants of each other in the true sense, rather than merely by co-incidental visualsimilarity.

Since, having such labels is a realistic possibility and the corresponding labels lookalmostexactlyalike,NBGPhasproposedthemasblockedvariants.

NBGPacknowledgestheconcernthatthisshapeisquitegenericandmayhaveparallelsin other scripts not under its ambit.However, as NBGP does not have any exposureaboutactualusageof thosecharacters in thoseparticularscripts,NBGPdesisted fromincludingthemintheanalysis.AsNBGPhasalreadyconsideredall therelatedscriptsunder the cross-script variant analysis, the similarity of the characters belonging toNBGP scripts with other scripts not under the NBGP ambit, may be of a mere co-incidentalvisualnature.

Additionally, this concern is not limited to these two characters but for all thecharacters inall the scriptsunder the scopeof theRootLGRprocedure.CarryingoutthisanalysiscanpracticallybedoneonlywiththeGenerationPanelsthatexistwhiletheNBGPisactive.ThisstillleavesoutthosescriptsoutofthescopewhichmaynothaveaGenerationPanel establishedyet.Hence,carryingout thisexercise inentirety isquiteimpracticable.Thisconundrumcanberesolvedifallthesuchcasesarehandledbythe"StringSimilarityAssessmentPanel"ofICANN.