Novelle,acollaborativeopensourcewritingtoolsoftware
FedericoGobbo
DICOM
21100,InsubriaUniversity
Varese,Italy
federico.gobbo@uninsubria.it
MicheleChinosi
DICOM
21100,InsubriaUniversity
Varese,Italy
michele.chinosi@gmail.com
MassimilianoPepe
DICOM
21100,InsubriaUniversity
Varese,Italy
massimiliano.p@gmail.com
Abstract
Inthispaperwediscussthenotionsofhy-
pertext,blog,wikiandcognitive mapping
inordertofindasolutiontothemainprob-
lemsofprocessingtextdatastoredinthese
forms.Weproposethestructureandarchi-
tectureof Novelle as a new environment
to composetexts. Its flexiblemodelal-
lows the collaborationfor contentsanda
detaileddescriptionof ownership. Data
are storedin a XMLrepository, so as to
use the capabilitiesof this language. To
developquicklyandefficientlywechoose
AJAXtechnologyover theRubyonRails
framework.
1 Introduction
Computationallinguistsare facingthe explosion
of new formsof writingas a massphenomenon.
Tellingpersonalandcollaborativestoriesthrought
webtechnologiesisknownundertheetiquettesof
‘blog’and ‘wiki’. It thereforebringsnew chal-
lengestothefieldofnaturallanguageprocessing.
Wearetryingtoaddressthembyrenderingexplic-
itly the structureof thesenew formsof text in a
way suitableforlinguisticcomputation.In order
to do so, wearebuildingan opensourcewriting
toolsoftware,calledNovelle.
1.1 HypertextasaNewWritingSpace
Bolter(1991)was the first scholarwhostressed
theimpactofthedigitalrevolutiontothemedium
of writing. Termsas ‘chapter’,‘page’or ‘foot-
note’simplybecomemeaninglessinthenewtexts,
or they highly change their meaning. When
Gutenberg inventedthe printingpressand Aldo
Manuzioinventedthe bookas we know it, new
formsofwritingsarose.Forexample,whenbooks
shouldn’t be copiedby handany longer, authors
tooktheadvantageandstartwritingoriginalbooks
andevaluation– i.e. literarycriticism– unlike in
theprevioustimes(Eisenstein,1983).Nowadays
the use of computersfor writinghas drammati-
callychanged,expeciallyaftertheirinterconnec-
tionviatheinternet,sinceat leastthefoundation
of the web(Berners-Lee,1999). For example,a
‘webpage’is moresimilarto an infinite canvas
thana writtenpage(McCloud,2001).Moreover,
whatseemstobelostistherelations,like thetex-
tureunderpinningthetext itself. Froma positive
pointofviewthesenewformsofwritingmayreal-
izethepostmodernistanddecostructionistdreams
of an ‘operaaperta’(openwork), as Eco would
defineit (1962).Froma morepessimisticone,an
authormayfeelto have lostpowerin thisopen-
ness. Henceforththe collaborative traitsof blogs
andwikis(McNeill,2005)emphasizeannotation,
comment,and strongediting. They give more
powerto readers,eventuallyfillingthe gap - the
so-calledactive readersbecomeauthorsas well.
Thissituationcouldmake new problemsriseup:
Whoownsthetext?Whichroleissuitableforau-
thors?We have toanalysethembeforepresenting
thearchitectureofNovelle.
1.2 Knownproblems
It is certainlytrue that wikisand blogsare new
formsof text. It is alsotruethatwehave already
mettheseproblemsinthefirstformofpurelydig-
ital texts – hypertexts. Now we are facingthe
samequestionduringprocessingtextsinblogsand
wikis.We considerhypertextsasparentsofblogs
andwikis.Ouraimistousetheanalysisofhyper-
textsforinterestinginsights,usefulforblogsand
wikistoo.
40
Followingthe exampleof Landow (1994),we
willcalltheautonomousunits of a hypertext lex-
ias (from ‘lexicon’),a word coinedby Roland
Barthes(1970).Consequently, a hypertextisa set
oflexias. Inhypertextstransitionsfromonelexia
toanotherarenotnecessarilysequential,butnav-
igational. Themainproblemsof hypertexts, ac-
knowledgedsincethebeginning,havebeentraced
asfollows(Nelson,1992):
• Theframingproblem, i.e. creatingarbitrary
closedcontexts of verylarge documentcol-
lections. When extractingsub-collections,
somelinksmaybecutoff.
• Comparingcomplex alternatives, i.e. to get
paralleloralternateversionsofthesamedoc-
umentin a simpleandeffective way, oneof
themaingoalofXanadu,theultimate“global
hypertext”dreamtbyNelson.
• Typologyoflinks, i.e.whenlinksbecometoo
many, weneedatypologyforlinks,avoiding
confusiontothereader/author.
• Version control, as the systemshouldkeep
trackof thehistoryof every document,sav-
ing differencesand pointingout correspon-
dencies.
We take fromwikisthe conceptof document
historyanditsconsequences.We considerit as a
goodapproximationof the‘versioncontrol’con-
ceptasshownabove.
Inwikiseverydocumentkeepstrackofitsown
history:creatinga documentmeanstostarta his-
tory, editinga documenttomove ahead,restoring
tomovebackontothehistorytimeline,destroying
a documentto stopthe historyitself. Moreover,
a sandboxis a temporaryview of a documentit-
self- i.e. a sandboxcannotcausea changeinthe
history(CunninghamandLeuf,2001). Figure1
showsthemodel.
creation
a very old
version
the document
history timeline
an old 
version
the current
version
the last
version
a restorean edit
destruction
sandbox
Figure1: Thedocumenthistorymodel
Historysnapshotsof thetimelinemaybe con-
sideredas permanentviews, i.e. views with a
timestamp. Consequently, exceptin the caseof
sandboxes, everychangein thedocumentcannot
be erased. Thismodelwillhave a strongimpact
ontheroleoflinksandontheunderpinningstruc-
tureofNovelleitself.
2 TheStructureofNovelle
Ouraimistocreateanopensourcehypertextmod-
elingsoftware,calledNovelle.‘Novelle’isanItal-
ian old-fashionedwordmeaning‘novels’,andin
Germanitmeans‘novel’too.ItresemblestheEn-
glishword‘novel’andtheFrenchword‘nuovelle’.
Webelievethatthisnameisclearlyunderstableto
every peopleeducatedin a European-basedcul-
ture,andthisiswhywehavechosenit.
Theemphasisonnarrativitytakes intoaccount
theuseofblogsaspublicdiariesontheweb,that
is stillthe maincurrentinterpretationof this lit-
erarygenre,or metagenre (McNeill,2005). Fur-
thermorewenoticedthatblogsandwikisarecur-
rentlysubjectedto osmosis,becausethey have in
commontheunderlyingcoretechnology. Soblogs
area literarymetagenrewhichstartedasauthored
personaldiariesor journals.Now they tryto col-
lectthemselvesinso-called‘blogspheres’.Onthe
otherside,wikisstartedascollectiveworkswhere
eachentryis notownedby a singleauthor- e.g.
Wikipedia(2005). Now personalwikitoolsare
arisingforbrainstormingandmindmapping.See
Section4forfurtheraspects.
2.1 TheProblemofOwnership
The maindifferencebetweenblogsand wikisis
in the ownershipof documents.Mostblogsfol-
low theannotationmodel, wherea singlelexiais
centraland the othersare comments,sometimes
in threads. Every lexia is authoredand changes
areminimal.Peopleprefercommentinginsteadof
editing.Theparadigmis“writeonce,readmany”.
On the contrary, in wikisno lexia is authored
andthereis no hierarchy betweenlexias. In fact
a documentisstilla setoflexias,buteverydocu-
mentis onlythe set of historicalversionsof the
documentitself. Generally, peopleavoid com-
menting,preferringto edit eachdocument. The
paradigmis “writemany, readmany” (Cunning-
hamandLeuf,2001).
Webelievethatownershiphasanimportantrole
and we do not want to forceour usersto take a
41
non-attributivecopyrightlicencetotheirwork.We
considertheCreativeCommonsmodelasthemost
suitableone to let eachauthorchoosethe rights
to reserve (Lessig,2004). Narrative writingsor
essaysarecreative worksandthey generallytreat
ownershipasauthorship,evenforthemostenthu-
siasticfellowsoffreeculture(Stallman,2001).
2.2 TheRepresentationofContext
InthestructureofNovellewearetryingto retain
authorshipandthecoreconceptofdocumenthis-
toryof wikisthrougha typologyof links,taking
whatwe considerthe best of the two worldsof
blogsandwikis.
In Novelleeachuserownshisownlexias,and
the relationsbetweenthem,i.e. links. Further-
more authorsare free to read and to link other
users’lexias. In otherwords,eachuserdoesper-
miteveryonetolinkitsownlexiasforfree,atthe
conditionthattheothersdothesame.Everyuser
mayrecallthelinklistoneachelement(e.g.asin-
gleword)ofhislexiasateverytime,buthecannot
destroythem.Lexiasmaybecommentedbyevery
user,buttheauthormayretainforhimselftheright
toedit.Thisdecisionhastobetakenwhena lexia
iscreated.
Ifa userletsotherseditsomelexias,hehasthe
righttoretain orrefusetheattributionwhenother
usershaveeditedit. Inthefirstinstance,theedited
versionsimplymovesaheadthedocumenthistory.
Inthesecondone,thelastuser,whohaseditedthe
lexia,mayclaimtheattributionforhimself.The
lexiawillbemarkedasaderivativeworkfromthe
originalone, and a new documenthistorytime-
linewillstart(seeFigure2). Authorsmaychoose
thisrightwiththeNo-DerivoptionoftheCreative
Commonslicences(Lessig,2004).
creation
the document
history timeline
an old 
version
the current
version
creation of a
derivative work
a new document
 history timeline
Figure2: Howtocreatederivativeworks
If nobodyclaimsthe documentfor himself,it
willfallinthepublicdomain.Thesetoflexiasin
thepublicdomainwillforma specialdocument,
ownedbya specialuser, calledPublicDomain.If
theauthorrefusesthepermissiontocreatederiva-
tive works,i.e. to edithis own lexias,usersstill
have therightto commenttheauthor’s work. So
astocometotermswiththisidea,weneeda con-
ceptinventedbyNelson(1992),i.e.transclusion.
the document
history timeline
the current
version
an other document
 history timeline
a freezed quotation
transclusion
Figure3: Howtransclusionworks
Rather than copy-and-pastecontentsfrom a
lexia, a user may recall a quotationof the au-
thor’s lexiaandwritea commentinthesurround-
ings. In doingso, the link list of the author’s
lexia willbe updatedwitha specialcitationlink
marker,calledquotationlink(seelaterfordetails).
Usually, the quotationwillbe ‘frozen’,as in the
momentwhereit was transcluded(seeFigure3).
Consequentlythetransclusionresemblesacopied-
and-pastedtext chunk,but thelinkto theoriginal
documentwillalwaysbeconsistent,i.e. neitherit
expiresnorit returnsanerror. Otherwisetheuser
whohastranscludedthequotationmaychooseto
keepupdatedthe linksto the originaldocument.
Thischoicehastobemadewhenthetransclusion
isdone.
the document
history timeline
an old
version
an other document
 history timeline
an up-to-date 
quotation
transclusion
the current
version
Figure4: Anup-to-datetransclusion
If so,thetranscludedquotationwillupdateau-
tomatically, followingthe historytimelineof the
originaldocument. For example,if the original
42
documentchangestopicfromstarsto pentagons,
the quotationtranscludedwill changetopic too
(seeFigure4).
2.3 ContentsandtheTypologyofLinks
Followingour modelof ownership,thereare at
least two categoriesof links: shallow linksand
deeplinks.Byshallowlinkswemeanvisuallinks
occurringinasinglecanvas,usuallyownedbythe
sameauthor. Thesewillrepresenticonicallythe
relationshipspaceof lexias,as explainedby Mc-
Cloud,talkingaboutwebcomics(2001).Theyare
particularlyusefulwhencomparingparallelver-
sionsof the sametext, e.g. digitalvariants(see
Conclusions).
We considera webpage,or bettera webcan-
vas,asa view of lexias, i.e. a groupoflexiasand
theirrelationsvisuallyshownwithshallow links.
Asetoflexiasisadocument. Everyauthorhasthe
rightto decidetherelationtypeofa setoflexias,
i.e. toforma document.Adocumentcanalsobe
consideredasacollectionofhistorytimelines,i.e.
thesetofrelatedlexiasandtheirversions.Theset
of documentsis thedocuverse, a wordcoinedby
Nelson(1992). We usethe worddocuverse,un-
likelytheoriginalsense,withthemeaningofaset
ofdocumentsownedbyasingleauthor.
Every documentcan be viewed withina web
canvas, but usersmay clickon a deep link and
so changeview. With deeplinkswe meanlinks
whichlettheuserchangeview, i.e. rearrangeel-
ementsin the web canvas for revealingshallow
linksbetweenlexias,belongingtothesamedocu-
mentornot.Thereforeawebcanvasmayshowre-
lationsbetweenviewsownedbydifferentauthors.
We considerquotationlinks,i.e. linkscreatedby
transclusion,as a specialkindof deeplinks. Au-
thorsmaycreatespecificviews addinglabelsto
links. Theset of labelswillforma typologyof
links,customizedby everyuserandeven shared,
ondemandofusers’desires.
Withourtypologyoflinks,weaimtosolve the
framingproblemas definedin Section1.2. We
wanttomodelviewsasdynamicobjects- thecre-
ationofcontextwillbestillarbitrary, butchanges
are very easily. We would also provide a user
facilityfor choosingthe right licencefor every
lexia,followingthemodelof Creative Commons
licences(Lessig,2004).
3 TheArchitectureofNovelle
We have consideredmany hypothesesin orderto
choosea firstlayerarchitectureto save a reposi-
tory. We useda multi-tiermodelbasedonXML.
Ourideaisbasedonmergingtogethersomeofthe
mostcommondesigntechniquesusedinblogsand
wikis. Recentlyprevious implementationtech-
niqueshavebeenstudiedfromtheirnewaspectsto
findinnovativewebtechnologies.Abasicscheme
of Novelle architectureis presentedin Figure5.
Thefirstlayeristhemostimportant.Itisbasedon
AJAX
Ruby on Rails
RDBMS
XML
DBMS / Filesystem
GUI
Figure5: Novelle:multi-tierarchitecture
aninfrastructureforstoringeffectivelydatarepos-
itoryinordertoobtainthebestperformances.We
have studiedtwo alternatives for the repository.
Ononesidewehave differenttechniquesto map
XMLtreesontoa databasemanagementsystem.
OntheothersidewemaymapXMLtreesdirectly
onafilesystem–seebelowfordetails.
ThesecondlayerisrepresentedbyXML.Mes-
sages,dataandmetadataareexchangedbetween
layersusingthecapabilityof thislanguage.This
allowstotreatdataandmetadataondifferentlevel
ofabstraction.
The Rubyon Rails(2006)framework permits
us to quicklydevelop web applicationswithout
rewritingcommonfunctionsandclasses.
We usedtheAsyncronousJavascriptAndXML
(or AJAX)paradigmto createthe graphicaluser
interface. AJAXfunctionletsthecommunication
worksasyncronouslybetweenaclientandaserver
throughasetofmessagesbasedonHTTPprotocol
andXML(Garrett,2005).
3.1 XMLrepository
We chosetouseXMLtreestostoretogetherdata,
metadata,messagesandtheirmeaningsbecauseit
hassomebenefits.Themostimportantis storing
43
XMLdata.TheotherbenefitsofanativeXMLso-
lutionare:thestoringwithoutmappingyourXML
tosomeotherdatastructurelikeobjects,classesor
tables;theneatnessof thestructure;theunderly-
ingtechnologyfromtheabstractlayertothephys-
icalonesisbasedonauniquestandard,widelyac-
ceptedby the community. Datamaybe entered
andretrievedasXML.Anotheradvantageisflexi-
bility,gainedthroughthesemi-structurednatureof
XMLandtheschemaindependentmodelusedby
mostofnative XMLdatabases.This is especially
valuablewhenyouhaveverycomplexXMLstruc-
turesthatwouldbedifficultorimpossibletomap
to a morestructureddatabase.At thistimethere
are not XMLdatabasesso stableto be usedinto
projectofthiskind.
Xindice(developedby ApacheGroup)proved
betterthanothers. ApacheXindiceis a database
designedfromthe groundup to storeXMLdata
orwhatis morecommonlyreferredtoasa native
XMLdatabase. It storesshortXMLdocuments
incollectionswithruntimegenerationofindexes.
UnfortunatelyXindiceseemsnottohavebeende-
velopedanymoresinceApril2004.
AnothernativeXMLdatabase,moreusableand
supported,iseXist. eXistisgrowingquicklyandit
implementssomefunctionalitiesof Xindice.The
standardssupportisnotcompletedandsomefunc-
tionsarecurrentlybeingrewrittendirectlyembed-
dedintothesoftware.Afterdoingmanytestsonit,
itrevealsworseperformanceswithrespecttoother
platforms,evenifitismorecompleteincompari-
sontotheothers.
Anyway our interestkeeps focusingon them
waitingfor the firststablereleaseeffectively us-
able in Novelle. We have consideredthe possi-
bilityto mapXMLtreesto relationalor object-
orienteddatabasemanagementsystemsthatsup-
portXML.WecanmapdirectlyanXMLtreeinto
amemorytreestructure,madeupwithclassesand
objects,withobject-orienteddatabases,aswecan
seein Ozoneproject(2006). Thelast stablever-
sion of Ozonewas releasedin 2004. The main
problemwithOzone- andwithothersOODBMS
- is the overheadrequestedto the memoryfor
storinga complex tree. On the otherside,many
RDBMSwithXMLsupportmapdirectlyanXML
treetoanentity-relationshipschema.Insomein-
stancesXMLtreesarestoredasBinaryLargeOb-
ject,or BLOB,intoonebig table. In othersitu-
ationsXMLtreesare parsed,splittedandfinally
storedin tableswhereattributes have the same
namesthanXMLnodes.
RonaldBourret(2006)mantainsandupdatesa
verycomprehensive listofnative XMLdatabases
onhiswebsite.
While we are waiting for a native XML
databasestableandusefulforourproject,wehave
decidedto getinspirationfromthecommonidea
usedinmanyblogsandwikis.Mostofthesearchi-
tecturesareusedto storemessagesin a structure
thatis similartoa directorytreesavedonfilesys-
tem.Oftenthisideaisonlydevelopedtopresentto
usersmessagesorganizedincollectionorderedby
time(e.g. blogs),but all the platformsarebased
on RDBMS.We have found in ourresearchonly
oneotherprojectinwhichmessagesarestoreddi-
rectlyon filesystem:the Gblog project(Gblog,
2005).Nobodyusuallyadoptthissolutionbecause
thesecurityof thewebsiteis lessstrong. In or-
dertorepresentmessagesarchivesthemostcom-
monstructureis thetriple../year/month/day/... In
our assumption,a messageis a history. There-
forea structureof thiskindworksverywellwith
ouridea.We aregoingtobuilda filesystemtime-
basedstructurein whichwecandirectlymapour
messagesi.e.ourhistories.Thisstructureisalsoa
tree. We can writealsoan XMLdocumentthat
mantainsan architectureschemewith some in-
dexes to speedup queries. Moreover, we store
withamessageanotherXMLdocumentrepresent-
ingallthepasthistory(i.e. thepaths)ofthemes-
sage.
So as to sum up, every time a user storesa
message,he hasto save thefirstXMLdocument
withthemessage,thensaves or updatesa second
XMLdocumentrepresentingits pasthistoryand
finallysavesorupdatesathirdXMLmessagewith
filesystemdirectorytree. Theoverheadon band-
with and net speedof this solutiondoes not let
usersnoticesignificantdifferences,eventhoughit
isnecessarytograntwritingpermissionstoevery-
oneontheentirerepository. HavinganativeXML
databasewillgive the advantageof savingXML
documentsin a rapid,neatandindicizedway, in
ordertobeabletoexecuteefficientqueriesonthe
repository.
3.2 eXtensibleMarkupLanguage
We choseXMLas languageand meta-language
becausewe neededto be able to save messages
with their meanings. Every lexia is saved with
44
sometagsandattributeswhichdescribeitsmean-
ing.Thepossibilityofstoringseparatelydatafrom
their representations lets a systemaccessmore
quicklyto a dataandextracttherequestedinfor-
mation.XMLis a W3Cstandardandthismakes
ourprojectreadyto be changedandextended,as
wellastobeconnectedwithotherapplicationsand
services(XML,2005).XMLwillbeusedtorep-
resentdata,metadata,linktyping,messagesand
pathsmap, and to exchangemessagesbetweeen
differentlayers.
3.3 RubyonRails
RubyonRails, orRoR,isa frameworkrichinex-
tensionsandlibrarieswithlicencessuitableforour
usage,inparticularXMLBuilderandgdiff/gpatch.
Thefirstlibraryoffersasetofclasseswhichallows
to generateXMLcodein a simpleway (Builder,
2006). Gdiff/gpatchlibraryis an implementation
ofthegdiff protocol,thatcreatesapatchfromtwo
files and then a new file fromone of the previ-
ousfilesandthepatch(Gdiff, 2005). Usingthis
librarywe are goingto be ableto storethe his-
toryandthelastversionin an easyway andsav-
ingspace.Creatinga documentis thereforea se-
quenceofpatches.Storingworksinthesameway,
thatis executinga gdiff protocolandstoringthe
new patch. Movingacrossthe documenthistory
meansretrievinganumberofpatchcommandsun-
tilyoureachthedesiredversionofthedocument.
Rubyon Rails does not supportnative XML
databasesat this time,thereforein our architec-
turetherewillbeprovisionallyarelationalDBMS
dedicatedto RoR,whichhadno problemwitha
filesystemrepository.
3.4 AsyncronousJavascriptAndXML
AJAXis nota technologyinitselfbuta termthat
refersto the use of a groupof technologiesto-
gether, inparticularJavascriptandXML.Inother
wordsAJAXis a webdevelopmenttechniquefor
creatinginteractivewebapplicationsusingacom-
binationof XHTMLand CSS, DocumentOb-
jectModel(orDOM),theXMLHTTPRequestob-
ject(Wikipedia,2005).
AJAX paradigmhas been recently defined,
whensomeonehas rediscovered a simplefunc-
tionoriginallydevelopedbyMicrosoftasActiveX
control.Thisfunction,namedXMLHTTPRequest
lets clientsask servers for some particulardata
usingasyncronoushandshake. In thisway users
cancontinueusingwebapplication(typicallyfill-
ing web forms)while the client and the server
exchangedata and messages. Otherdevelopers
have publisheda concurrentversionof thisfunc-
tion for other browsers than InternetExplorer,
like Mozilla/Gecko, Operaand Safari. Theweb
pagesbuildedwiththis technologygive the im-
pressionto have dynamiccontent.Importantex-
amplesbuildedwithAJAXparadigmareGmailby
Google,Writely, Kiko, Webnote,Meebo. Using
AJAXto developwebapplicationsandwebser-
vicesneedssomeattention.FirstofallJavascript
mustnotbedisabledinbrowsers.Itisalsoneces-
sarytopayattentiontoestimatecorrectlythetime
spentin exchangingmessagesbetweenclientand
server so to exploitthe goodcapabilitiesgained
withAJAX,fullysupportedby andintegratedin
RubyonRails.
3.5 Accesspoints
We are goingto add to every view of Novelle a
searchenginethatreturnsa list of meaningsand
a set of linkbetweenthem. Theselinksare rep-
resentedinourprojectwithimages.Everyimage
is itselfa mapthattheusercansurfand/oropen
toincreasedetailslevel. Whentheuserhasfound
the message,he can accessto it simplyclicking
onit. Anusercancommentormodifyeverylexia,
if theseactionsaregrantedbytheoriginalauthor,
as explainedabove. Userscan createnew links
betweenlexias and they can describewhatkind
of linkthey intendto createthroughappropriate
linktype.Thesemodificationsarestoredusingthe
documenthistorymodelof Novelle throughfol-
lowingpatch.
4 RelatedWorks
Themainsourceof Novelle arewikisandblogs.
While wikis have spread from a detailed de-
sign(CunninghamandLeuf,2001),unfortunately
blogs have not been designedunder a model.
So we have testedand comparedthe mostused
toolsavailableforblogging:Bloggers,WordPress,
MovableTypeandLiveJournal.
Generallyspeaking,we find that the personal
public diary metaphorbehind blogs (McNeill,
2005)maybringto an unsatisfactoryrepresenta-
tionofthecontext.Theonlywaytoretrieveinfor-
mationis througha searchengineor a calendar,
i.e. thedateofthe‘post’– a lexiainthejargonof
bloggers.
Moreover, we usesomenew webapplications
45
to take andsharenotesor to browsereveryone’s
bookmarks,e.g. del.icio.us. Mostly, theseweb
applicationsorientedto writinggive a strongem-
phasison collaborationandsharing. Thisled us
torethinkownershipandtousetheCreativeCom-
monsmodeltodesignthecontentsofNovelle.
Finally, we noticed that personal wikis are
used for storing cognitive maps of individuals
andbrainstorming.Thisusewas alreadythought
bythefoundersof wikis(CunninghamandLeuf,
2001),butithasnotbeenwidelyexploredinprac-
tics,asfarastheauthorsknow. However, thisdi-
rectionof workis notactuallynew - conceptand
mindmapping,thetwomainparadigmsforcogni-
tivemaps,havebeenusedforseveralyears.
Conceptmappinghasbeenusedat leastin ed-
ucationfor over thirtyyears,in particularat the
CornellUniversity, wherePiaget’s ideasgave the
rootstotheassimilationtheorybyDavidAusubel.
Verybriefly, conceptmapsshowtherelationships
betweenconceptslabellingbothnodesand arcs.
Everyarcalwayshasa definitedirection,i.e. arcs
arearrows(Novak,1998).
In contrast,mindmapsspreadfroma centre,
withbranchesradiatingout. Furthermore,mind
maps,asthoughtandcopyrightedbyTonyBuzan,
can label only nodes, not arcs. The resulting
shapeof mindmapsis sometimessimilarto neu-
rons’(Buzan,2000).
Wehavetestedbothconceptandmindmapping
software tools, availablefor freeor in a trialpe-
riod. In particular, CmapTools software is cur-
rentlyusedattheCornellUniversityandit isfree
as a client. It mayrunonCmapServers,andit is
a verygoodwaytosharetheknowledgestoredin
cognitive maps.Unfortunately, it doesnotcollect
datain a formatsuitablefortheweb,andit does
notpermittoviewconceptsacrosscognitivemaps
ownedby differentusers(Tergan, 2005). More,
conceptmapsrequirea learningcurve very high
whenstartedtobeused,atleastinourexperience.
Onthecontrary, mindmapsarebyfarmoreintu-
itive.
Thereare a lot of mindmappingtools,which
are clonesof MindJetMindManager, the official
software for Buzan’s mindmapping. The mind
mappingtoolwewerelookingforshouldhavehad
anopensourcelicence,usedaformatfordatastor-
agesuitablefortheweb,andbeencross-platform.
Infact,Freemind,astheclosestapproximationof
ourneeds(Mueller, 2000),succeededin running
on the three majoroperatingsystemsavailable,
withoutsensibledifferences.
Figure6: OurfreemindmapforNovelle
Evenif welike theideabehindmindmaps,we
needtohave a multiauthoredenvironment,where
arcs may be labeled. In otherterms,the centre
of themapshouldchangeaccordingto theuser’s
desire.Thatis why wethoughtaboutwebcanvas
as documentviews. If weconsiderdocumentsas
freemindmaps,thenodeswillbe lexiasandthe
arcswillbelinks.
Apartfrom wikis, blogs, and cognitive map-
ping,wewerealso inspiredbytheexperiencesof
earlyhypertext writingtools,in particularInter-
mediaand Storyspace. In fact, they were used
expeciallyin academicwritingwith some suc-
cess. Intermediais no moredevelopedand no-
bodyof us hadtheopportunityto tryit (Landow,
1994).StoryspaceiscurrentlydistributedbyEast-
gate(2005),andwehave useditfora time.How-
ever, inouropinionStoryspaceis a productofits
time and in fact it isn’t a web application. Al-
thoughit is possible to labellinks,it lacksa lot
offeaturesweneed.Moreover, nohypertextwrit-
ingtoolavailableisreleasedunderanopensource
licence.WehopethatNovellewillbridgethisgap
- wewillchoosetheexactlicencewhenourfirst
publicreleaseisready.
We arepersuadedthatthereis nocontradiction
incollaborativemindmappingandacademicwrit-
ing. Maybeit is notby chancethatEastgatehas
alsoreleaseda “personalcontentmanagementas-
sistant”(Eastgate,2006).Ourpurposeis tobring
back again collaborative writingand free brain-
storming,asitshouldbe.
46
5 ConclusionsandFurtherWorks
We arecurrentlydevelopinga prototypeof Nov-
elle.WearguethatthemodelunderNovellewould
be an explicitrepresentationof thecontext anda
clearmodelforthecontents.Oneofthemainap-
plicationof oursoftwareis naturallanguagepro-
cessing.Wearegoingtotestitexpeciallyondigi-
talvariantsofliterarytexts.
Acknoweledgements
We want to acknowledge Dr. Marco Benini,
Dr. AlbertoTrombettafortheirdeepinsightsand
GabriellaCancianifor havingreviewedthe final
draftofthispaper.

References
RolandBarthes.1970.S/Z. EditionsduSeuil,Paris.
Tim Berners-Lee.1999. Weavingthe Web. Harper,
SanFrancisco.
JayDavidBolter. 1991.WritingSpace:theComputer,
Hypertext,andtheHistoryofWriting. ErlbaumAs-
sociates,Hillsdale,N.J.
RonaldBourret.2006. XMLDatabaseProducts. Url:
http://www.rpbourret.com/xml/. Retrieved the 3rd
ofJanuary.
Builderlibrary.for example 2006. Project: Builder.
Provide a simpleway to create XMLmarkupand
datastructures. Url: http://builder.rubyforege.org/.
Retrievedthe4th ofJanuary.
Tony BuzanandBarryBuzan. 2000. TheMindMap
Book. BBCWorldwideLimited,London.
WardCunninghamandBoLeuf. 2001. TheWikiWay
-QuickCollaborationontheWeb. Addison-Wesley,
Boston.
Eastgate 2005. Storyspace. Url:
http://www.eastgate.com/storyspace. Retrieved
the31st ofDecember.
Eastgate 2006. Tinderbox. Url:
http://www.eastgate.com/tinderbox. Retrieved
the2nd ofJanuary.
UmbertoEco. 1962.Opera aperta. Bompiani,Milan,
Italy.
ElizabethL. Eisenstein.1983. ThePrintingRevolu-
tionin EarlyModernEurope. CambridgeUniver-
sityPress,Cambridge,UK.
Jesse James Garrett. 2005. Ajax: A New
Approach to Web Applications. Url:
http://www.adaptivepath.com/publications/essays/
/archives/000385.php. Retrieved the 22nd of
December.
Gblog2.0. 2005. Gblog 2.0. Blog, reloaded. Url:
http://gblog.com/.Retrievedthe27th ofDecember.
Gdiff/Gpatch library. 2005. Gdiff/Gpatch. An
implementationof the W3C gdiff protocol. Url:
http://ruby.brian-schroeder.de/gdiff/. Retrieved the
28th ofDecember.
GeorgeP. Landow 1994. Hypertext 2.0.TheConver-
genceof Contemporary CriticalTheoryand Tech-
nology. TheJohnsHopkinsUniversityPress,Balti-
more,Maryland.
LawrenceLessig2004.FreeCulture: HowBigMedia
UsesTechnologyandtheLawtoLockDownCulture
andControlCreativity. PenguinBooks.
ScottMcCloud.2001. UnderstandingComics. Para-
doxPress,NY.
LaurieMcNeill. 2005. GenreUnderConstruction:
TheDiaryontheInternet.Language@Internet, 1.
Joerg Mueller. 2000. FreeMind. Url:
http://freemind.sourceforge.net. Retrieved the
31st ofDecember2005.
TheodorHolmNelson.1992.LiteraryMachines90.0.
Muzzio,Padua,Italy.
JosephDonaldNovak. 1998. Learning, Creating,
andUsingKnowledge: Concept MapsAs Facilita-
tiveToolsin SchoolsandCorporations. Lawrence
ErlbaumAssociates.
Ozone DatabaseProject. 2006. Ozone Database
Projectopeninitative.Url:http://ozone-db.org/. Re-
trievedthe03rd ofJanuary.
Ruby on Rails. 2006. Ruby on Rails.
Web developement that doesn’t hurt. Url:
http://www.rubyonrails.org/. Retrieved the03rd of
January.
RichardM.Stallman.2001.FreeSoftware, FreeSoci-
ety: SelectedEssaysofRichard M.Stallman. GNU
Press,Cambridge,Massachusetts.
Sigmar-OlafTerganandTanjaKellers. 2005. Knowl-
edge AndInformationVisualization:Searchingfor
Synergies. Springer, Berlin.
Dave ThomasandDavidHeinemeierHansson.2005.
AgileWeb Developmentwith Rails- A pragmatic
guide. PragmaticBookshelf.
Wikipedia. 2005. Wikipedia. From
Wikipedia, the free encyclopedia. Url:
http://en.wikipedia.org/wiki/Wikipedia. Retrieved
the31st ofDecember.
XML, eXtensible Markup Language. 2005.
Extensible Markup Language (XML). Url:
http://www.w3.org/XML/. Retrieved the 27th
ofDecember.
