AN EVENT STRUCTURE BASED COORDINATION

MODEL FOR COLLABORATIVE SESSIONS

Kamel Barkaoui

, Chafia Bouanaka

and

José Martín Molina Espinosa

CEDRIC Research Laboratory, CNAM, Paris, France

LIRE Laboratory, Mentouri University, Constantine, Algeria

Tecnológico de Monterrey, Campus Ciudad de México, Mexico

Keywords: Session Management, Collaboration, Event Structures, Computer Supported Collaborative Work.

Abstract: Distributed collaborative applications are characterized by supporting groups’ collaborative activities. This

kind of applications is branded by physically distributed user groups, who cooperate by interactions and are

gathered in work sessions. The effective result of collaboration in a session is the production of

simultaneous and concurrent actions. Interactions are fundamental actions of a collaborative session and

require being coordinated (synchronized) to avoid inconsistencies. We propose in the present work an event

structure based model for coordination in a collaborative session, making possible interactions between

participants and applications in a consistent way. The proposed model describes interdependencies, in the

form of coordination rules, between different actions of the collaborative session actors.

1 INTRODUCTION

Distributed collaborative applications are

characterized by supporting activities of physically

or virtually distributed groups which cooperate by

interactions. These activities are performed by user

groups and are gathered in work sessions. These

sessions constitute the basic units of collaboration.

The effective result of collaboration in a session is

the production of concurrent actions carried out

during definition and execution phases of the

session. These actions require being coordinated in

order to avoid inconsistencies.

Collaborative sessions represent a space in which

various entities, such as participants, applications

and data interact. Collaboration goal is to aid

participants in the achievement of their tasks by

giving them the possibility to exchange knowledge

and information. These interactions imply that

various actions of session actors tend to achieve a

common goal. That is why actions might be

complementary and assigned to actors with respect

to their skills and /or functionalities. Interaction

result, carried out during a collaborative session,

strongly depends on the order in which they are

executed. Indeed, interactions can be destructive if

they are not coordinated.

Clearly, collaboration is the reachability of common

goals of the underlying session. Reaching such goals

depends on the respect of partial ordering between

actions executed by all actors. Such order is depends

on certain dependencies between actions. Using

event concept, event occurrences obey to some

coordination rules. Hence, we need formalism able

to capture coordination between events. Ultimately,

we require a formalism offering tools to express

event occurrences with respect to causality, conflict

or concurrency. Event structures (Winskel, 1987),

(Winskel, 1992) are an adequate formalism to

express mathematically these relations on sets of

events. In this work, we present a model for

collaborative sessions, based on the specification of

interactions between actors during a session. The

proposed model allows formal specification of

different causal dependencies between events:

precedence, inhibition, and release. This paper is

organized as follows. We begin by a brief

presentation of functional description of services

implied in collaborative session management. In

section 3, we recall essential aspects of event

structures. In section 4, we introduce a new

coordination model for collaborative session

management, where basic coordination rules are

well defined. In section 5, we present a mapping

from event structures to Petri nets in order to exploit

137

Barkaoui K., Bouanaka C. and Martín Molina Espinosa J. (2009).

AN EVENT STRUCTURE BASED COORDINATION MODEL FOR COLLABORATIVE SESSIONS.

In Proceedings of the 11th International Conference on Enterprise Information Systems - Information Systems Analysis and Speciﬁcation, pages

137-143

DOI: 10.5220/0001987101370143

 SciTePress

Petri nets rich panel of tools for system verification.

In section 6, we compare the proposed approach to

other Models of collaborative sessions. We conclude

by presenting directives for future work.

2 AN OVERVIEW OF CSCW

Computer Supported Cooperative Work (CSCW)

refers to the study field concerned with design,

adoption, and use of groupware (Ellis, 1994).

Despite its name, this study field is not restricted to

issues of cooperation or work but also examines

competition, socialization and play. Computer

Supported Cooperative Work is considered as a new

research field involved in exploring a wide range of

issues concerning cooperative work arrangements

and support via information technology.

2.1 Actors of a Collaborative Session

The term session management refers to the process

of starting, stopping, joining, leaving, and browsing

collaborative situations across a network (Molina

Espinosa, 2003b). Generally, we distinguish three

types of users:

• A session participant: is a user taking part in a

collaborative session and using applications

assigned to session in progress.

• The session chair: is a participant who moreover

controls session running (right to speak, tools

access), invites participants, initializes, opens,

and finishes the session.

• Session administrator: is responsible of defining

and planning the session. He establishes

participants list including session chair. These

participants are selected according to their roles,

their competencies, and their skills.

2.2 Functionalities of a Collaborative

Session

We can identify two essential phases in a

collaborative session: The first stage concerns

session preparation. It consists of defining and

updating participants list of the collaborative session

and corresponds to Responsibility and Management

Service (RMS). The second stage concerns session

management and takes care of actions coordination

during collaborative activities accomplishment. It is

realised by Session Management Service (SMS).

2.2.1 Responsibility and User Management

Service

RMS provides necessary functions to define

participants profile, groups, and sessions (Molina-

Espinosa, 2003a). It also provides information to

SMS service and actually used applications.

RMS is composed of three main activities:

Participants list definition, i.e., users intervening in

the collaborative session. Session group’s definition,

each group represents a set of participants, and

finally, establishment of active session list.

2.2.2 Session Management Service

SMS offers a set of mechanisms by which users can

initialize, find, open, close, join, leave and finish a

collaborative session. Basic functions of session

management are:

• Session Initialization: corresponds to necessary

objects creation by session chair. These objects

are used to perform coordination actions during

session evolution.

• Session Announcement: consists of inviting

potential participants to a session. It also

includes participants reply (Acceptance or

Refusal).

• Session open and join: A session is opened by

the session chair allowing participants to join it

through their work stations.

• Session leave: allows participants to disconnect

from a session.

• Session closure: allows session chair to

terminate the collaborative work.

• Session termination: includes destruction of

session objects, created during session

initialization. This task is performed by session

chair.

3 EVENT STRUCTURES

Generally, it is immaterial to analyze the precise

places and times of event occurrences in a

distributed computation (Winskel, 1992). Significant

events and how their occurrence depends on the

previous occurrence of other events are more

important. Hence, distributed computations are

considered as a set of event occurrences doted with a

causal dependency relation between them. Defining

a partial order between event occurrences is then

reasonable. To model non-determinism and express

how some event occurrences rule out the occurrence

of others, a conflict relation between events is

ICEIS 2009 - International Conference on Enterprise Information Systems

138

defined. The set of events with causal dependencies

and conflict relations are called event structures.

Event structures are abstract descriptions of a

computation focusing on significant events of the

computation and describing the possible ways that

computation could follow.

Definition 1 (Winskel, 1987). A labelled event

structure is a quadruple

),#,,,( ALEES ≤=

where:



is a finite set of events;



EE ×⊆≤ is a partial order relation called

causal dependency relation which satisfies

}{

edEdEe ≤∈∈∀ / is finite;



EE ×⊆# is an irreflexive and symmetric and

relation, called conflict relation, which

satisfies

∀

e, e’, e”

∈

E : e

≤

e’ and e # e”

⇒

e’ #e”;

 A is a set of actions;



AEL →: is a labelling function on events,

associating to each event e the corresponding

action L(e).

Concurrency or causal independency between

events is a derived notion. Two events e, e’ are

concurrent and noted e co e’ if and only if:

))'#()'()'(( eeeeee ∨

≤

∨≤¬

(1)

A notion of computation state of an event structure

can be defined (Winskel, 1992). Taking a process

computation state as a set X of previously occurred

events in the computation. We expect that:

• If an event has occurred this implies that all

events on which it causally depends have

occurred too.

• No two conflicting events can occur together in

the same computation.

Definition 2 (Winskel, 1992). Let

),#,,,( ALEES ≤=

be an event structure.

Defining its configurations

()

LED ,#,,≤ consists

of those subsets X

⊆

E which are:

 Conflict free:

)'#(,', eeXee

∈∀

(2)

 Left Closure:

XeXeandeeEee ∈⇒∈≤∈∀ ')()'(,',

(3)

These two conditions allow us to construct

consistent configurations (executions).

4 A FORMAL MODEL FOR

COLLABORATIVE SESSIONS

Collaboration aim is the attainability of common

goals pursued by a session. Reaching such goals

depends on respecting partial orders defined

between actor’s actions and induced by

dependencies between actions. Using event concept,

event occurrence might obey to some coordination

rules. So, we need formalism able to capture

coordination between events and offers necessary

tools to express event occurrences with respect to

causality, conflict or concurrency. To model the

collaborative session by an event structure, we begin

by identifying the set of inherent events and their

relationships. During a session, actors perform some

actions corresponding to some event occurrences.

Hence, we define first the actors set; actions set, and

define the concept of event thereafter.

4.1 Actors Set

We use T

Basic

to denote the set of collaborative

applications,

{

}

ENSTMSGCTSMTT Basic ,,,

(4)

• SMT is Session Management Tool. It allows

session configuration, participants invitation,

and session control;

• GCT, for Group Conferencing Tool, allows

direct discussion between a group of

participants;

• TMS is Tool Management Service;

• ENS, Event Notification Service, informs

participants about actions made by other ones.

Basic

denotes the list of session participants.

Actor set is defined by N

Basic

= P

Basic

∪

Basic

. One

participant in a session has the role of session chair.

{

}

{

}

nipiirSessionChaP Basic ≤≤∪

(5)

4.2 Actions Set

Actions set A

Basic

contains all actions performed

during a session. A

, A

Chair

, A

denote respectively

actions performed by all participants, session chair

and application tools. We have:

TChairpBasic AAAA ∪∪

(6)

where:

{

}

messageacceptleavejoinA P ,,,=

(7)

AN EVENT STRUCTURE BASED COORDINATION MODEL FOR COLLABORATIVE SESSIONS

139

A participant connects himself to the collaborative

session by executing a join action. He leaves the

session by a leave action and receives or sends a

message to another participant via message action.

Finally, a participant can accept the session Chair

role by performing an accept action,

{

}

Chair pA A grant=∪

(8)

grant action consists of passing the session Chair

role to another participant,

ENSTMSGCTSMTT AAAAA ∪∪∪=

(9)

includes a variety of actions made available by

services and applications used during a collaborative

session. Actions ensured by Session Management

Service are create, delete, open, close and invite.

A videoconference application, noted GCT, allows

enabling or disabling a videoconference associated

to a collaborative session by performing the

enable/disable actions respectively. Management

Groupware Service, TMS, allows starting and

stopping groupware applications by carrying out

start and stop actions. Event Notification Service,

ENS, allows sending and receiving events through

push and pull actions execution.

4.3 Event Definition

An event e = (n, a), where (n, a) is a pair belonging

to N

A, is said to be occurred if action a is

performed by actor n. We need the two following

functions: Act and Mgt.

NEAct →:

, determines an event actor. It is

defined by:

),()(/)( aneLAaneAct =∈∃⇔=

(10)

AEMgt →:

, determines an event occurrence

associated action. It is defined as:

),()(/)( aneLNnaeMgt =∈∃⇔=

(11)

4.4 Main Relationships

Dependencies between events can be categorized in

three classes: precedence, inhibition and trigger

relations. They are significant for coordination rules,

between collaborative session actors, definition.

4.4.1 Precedence Relation

Precedence relation expresses the fact that executing

some action a’ (event e’) might be preceded by

eventual execution (occurrence) of another action a

(event e).

CeCCDCCeeed

∈⇒⎯→⎯∈∀≡ ':',)',(Pr

(12)

4.4.2 Inhibition Relation

Inhibition relation forbids the execution (occurrence)

of an action a’ (the associated event e’) if action a

(event e) is already executed (occurred).

CeCCDCCeeInhib

∉⇒⎯→⎯∈∀≡ '':',)',(

(13)

4.4.3 Trigger Relation

To express immediate causality between events, we

define trigger relation. That is, an event occurrence e

follows immediately the occurrence of event e’.

Formally, we write:

'/"';')',(

CCDCCCDCeeTrig

⎯→⎯∈∃⇒⎯→⎯∈∀≡

(14)

Finally, we define a session state concept. It can be

viewed as a configuration in the corresponding event

structure.

Definition 3. Let

),,#,,( ALE

≤

be an event

structure modelling the collaborative session. If D

denotes the set of its finite configurations, then:

CeCeDCee ∈⇒∈

∈

∀

⇔

≤

':'

(15)

CeCeDCee ∉⇒

∈

∀

⇔

':'#

(16)

'ecoe

andCeDCC ∈

∈

∃

⇔

:',

DCCandCe ∈∪

∈

(17)

Mapping from a session state to another is due to an

action execution by an actor. It can be considered as

a transition from configuration C to configuration C’

when event e occurs. We write:

}{'' eCCandCeCC

∪=∉⇔⎯→⎯

(18)

4.5 Modelling Session Actors

The dynamic behaviour of a collaborative session

requires expressing all constraints to respect during a

session evolution. These constraints are called

coordination rules and are expressed as follows:

Let e

, …,e

be events, p a participant, and t be an

application; then we have:

4.5.1 Event Structure Associated to a

Participant

Event structure associated to a participant i is

defined by ES

= (E

≤

, L

) where:

• E

is the set of events occurring due to actions

performed by a participant, and equals to {e

, e

}. These events are labelled

ICEIS 2009 - International Conference on Enterprise Information Systems

140

respectively (p, join), (p, leave), (p, accept), (p,

message).

• Causal dependency expresses the fact that :

Any participant can receive or send a message

only if he was already connected, i.e. (p, join)

≤

(p, message). Any participant can disconnect

from a session only if he was already connected,

i.e. (p, join)

≤

(p, leave). Any participant can

accept Session Chair only if he is a member of

the group (connected to the session), i.e. (p,

join)

≤

(p, accept). So, we have:

≤

= {(e

, e

), (e

, e

), (e

, e

)}

(19)

• Conflict relation expresses inconsistencies that

might be forbidden : Any participant cannot

receive or send messages as soon as he

disconnects from a session, i.e. (p, leave)

(p,

message)

= {((p, leave), (p, message))}

(20)

A Session Chair is a particular participant. He has

the ability to grant this role to another participant.

Thus, ES

chair

= (E

chair

≤

chair

, L

chair

) where:

• E

chair

= E

∪

chair

} / L(e

chair

) = (Chair, grant)

•

≤

chair

≤

∪

{( e

chair

)} / L(e

) = (p, leave)

A Session Chair cannot leave the session before

granting his role to another participant. :

chair

4.5.2 Event Structure Associated to Session

Management Tool

SMT associated event structure is defined by ES

SMT

= (E

SMT

≤

SMT

, L

SMT

) where: #

SMT

= ∅

• E

SMT

is composed of five events e

SMT1

, e

SMT2

SMT3

, e

SMT4

, e

SMT5

, labelled respectively (SMT,

create), (SMT, invite), (SMT, open), (SMT,

close), (SMT, delete).

• ≤

SMT

= {( e

SMT1

, e

SMT2

), ( e

SMT2

, e

SMT3

( e

SMT3

, e

SMT4

), ( e

SMT4

, e

SMT5

)}

4.5.3 Event Structure Associated to an

Application

Event structure associated to an application is

defined by ES

TMS

= (E

TMS

≤

TMS

, L

TMS

) where:

TMS

= {e

TMS1

, e

TMS2

}, labelled (TMS, start) and

(TMS, stop) respectively,

≤

TMS

= {( e

TMS1

TMS2

)}

and #

SMT

∅

4.6 Coordination Rules

Collaborative session global behaviour corresponds

to a parallel composition of individual behaviours

(expressed by event structures defined bellow) of all

actors coordinated by the following global rules to

avoid inconsistencies: Coordination rules for

participant admission and leave defines the

coordinated behaviour of all participants of a session

plus session management tool:

• Rule 1: Any participant admission to a session

is allowed only if he was invited to that session:

∃

∈

SMT

, e

∈

, L(e

) = (SMT, invite)

∧

L(e

) = (p,join)

⇒

Pred (e

)

• Rule 2: Any participant admission to a session

is forbidden as soon as this session is closed :

∃

∈

SMT

, e

∈

, L(e

)= (SMT,close)

∧

L(e

) = (p, join)

⇒

Inhib(e

)

• Rule 3 : Any participant is automatically

disconnected from a session as soon as it is

closed:

∃

∈

SMT

, e

∈

, L(e

) = (SMT, close)

∧

L(e

) = (p,leave)

⇒

Trig(e

)

Coordination rule for session chair delegation is:

• Rule 4 : Any participant who accept the role of

Session Chair becomes so only if he is invited

to this role by the current chair:

∃

∈

chair

∈

, L(e

)=(Chair, grant)

∧

L(e

)=(p, accept)

⇒

Pred(e

)

Coordination rule for application integration to the

session is:

• Rule 5 : Any application is stopped as soon as

the session is deleted, i.e.

∃

∈

SMT

, e

∈

TMS

L(e

)=(SMT, close)

∧

L(e

)=(p,stop)

⇒

Trig(e

)

5 SYSTEM VERIFICATION

Model verification is a crucial phase in system

specification. It allows detecting eventual

inconsistencies in the proposed model. Since Petri

nets offer a variety of tools for analyzing system

properties, we propose to exploit the mapping

(Nielsen, 1981) between event structures and a sub-

class of Petri nets, condition/event nets, to verify

some properties of the event structure based model

for collaborative sessions. The mapping is

constructed at a categorical level through a

coreflection composed of two functors. First funtor

shows how to embed the category of event structures

into that of Petri Nets. The other funtor abstracts

away system implementation details, i.e. Petri Net

places, to bring it back to an event structure. In this

work, we are interested of the first functor, where an

event structure is identifiable to an occurrence net

ensuring causal and conflict relations defined in the

AN EVENT STRUCTURE BASED COORDINATION MODEL FOR COLLABORATIVE SESSIONS

141

event structure and obtained from execution of the

condition/event net defined as follows:

Definition 4 (Nielsen, 1981). A condition/event net

is a quadruplet (B,E,F,M

) where:

• B is a set of non null conditions,

• E is a disjoint set of events,

• F is set of (B×E)

∪

(E×B) called causal

dependency relation,

• M

is a non empty set of conditions, called

initial marking.

The Petri net satisfies the following restrictions:

•

∀

∈

∃

∈

B / F

b,e

0 and

∀

∈

∃

∈

B / F

e,b

no isolated events

•

∀

∈

B, [M

≠0 or (

∃

∈

E, F

e,b

≠0) or (

∃

∈

b,e

≠0)]; no isolated conditions

Definition 5 (Nielsen, 1981). Let E=(E,≤,

) be an

event structure, the corresponding occurrence net is

defined by N(E)=(B,E,F,M) such that :

• M={(Φ,A) / A

⊆

E and (

∀

a,a’

∈

A, a(

#∪

)a′)}

• B=M

∪

{(e,A) / A

⊑

E and e

∈

E and (

∀

a,a′

∈

#∪

)a′)and (

∀

∈

A, e<a)}

• F={(e,(c,A)) / (e,A)

∈

∪

{((c,A),e) / (c,A)

∈

and e

∈

6 RELATED WORK

Several models have been established for session

management, such as CONCHA, GMS, Mediaspace

and Intermezzo. CONCHA model (CONference

system based on java and Corba event CHAnnels

service) presented in (Orvalho, 1999) is a

supervising authority of conferences based on

CORBA events service. Services are implemented in

Java and support reliable multicast communications

for data transfer and information control. CONCHA

includes two essential services: supervising

conference authority and a multipoint

communication service. In (Wilde, 1996), Group

Management System for Distributed Multimedia

Applications (GSM) model is presented. GSM is

constitued of user agents and system agents. User

agents are components being integrated in the group

communication platform. System agents function is

to manage distributed directories providing

distributed databases to all user agents. In (Roussel,

1997), there is an inefficiency to collaborate in

traditional media space systems because

possibilities, to express coordination and actions,

have been pointed out. Roussel proposed a new

model, called Mediaspace, to handle collaboration.

This model is rooted in a multi agents approach.

Roussel characterized an agent by four fundamental

properties: persistence, ability to/for communication,

autonomy, and reactivity. In (Edward, 1994), a

session management model based on sharing users

and activities information is presented. Activities

information include current tasks details, active tasks

details (e.g., connected users), location of

applications or tasks, and objects associated with

these tasks. While CONCHA, GMS and Mediaspace

models are centered on distributed entities (agents or

components) integration, Edward’s model is based

on shared objects. All these models do not consider

interdependencies management between participants

and applications. Current session managers offer few

possibilities for coordination rules definition. In the

literature, there exist several works centered on the

definition of relations among participants,

applications and information (Rodriguez, 2002),

(Tata, 2002). Rodriguez et al. (Rodriguez, 2002)

describe application architecture to determine data

flows between producer-consumer components. Tata

(Tata, 2002) defines coordination policies based on

data access and synchronization contracts

established between members of a virtual team. This

model is centered on role management and activity

synchronization. It also supports inference of access

rules across a set of basic data. We have centered

our model on the use of participants and application

events without depending on data aspect. Our

approach consists in specifying a variety of

dependency relationships during cooperative session

execution. The proposed model is based on event

structures, giving more clarity and formality to

application specification. There are many

similarities between the proposed model and

Espinosa et al. (Molina-Espinosa, 2003b) one in

defining dependency relationships. However, the

major difference is the formal model used to specify

cooperative sessions and coordination rules. While,

Espinosa et al. used the Labeled Partial Orders

(LPO) for collaborative session definition and First

Order Formulas (FOL) to specify properties

corresponding to coordination rules. We have used

Event structures to specify both collaborative

session and coordination rules. This allows us to

discard ambiguity that appears in (Molina-Espinosa,

2003b) model.

7 CONCLUSIONS

In this paper, we have formally modelled

collaborative sessions using event structures. The

major benefit of such a model is the clarity of

ICEIS 2009 - International Conference on Enterprise Information Systems

142

various aspects in session management. In fact,

event structures give a theoretical foundation to our

model, allowing formal verification of some

properties. The work may be extended by proposing

an automatic mapping to other models and

formalisms such as Petri nets. Our current research

is to specify coordination rules by an extension of

event structures to domain event structures in order

to manage multiple instances of the same tool.

Mapping the specification expressed within event

structures to operational framework is an other

direction of future work.

REFERENCES

Edward W. K., 1994. "session Management for

Collaborative Applications". Proceedings of the ACM

Conference on Computer Supported Cooperative

Work (CSCW'94), Chapel Hill, NC.

Ellis C. A., Wainer J., 1994. "A Conceptual Model of

Groupware". CSCW'94 Conference on Computer

Supported Cooperative Work, pp. 79-88, ACM,

Chapel-Hill, North Carolina, USA, .

Molina-Espinosa J. M., Fanchon J., Drira K., 2003a. "A

logical Model for Coordination Rule Classes in

Collaborative sessions", Rapport LAAS N03132,

IEEE International Workshops on Enabling

Technologies: infrastructure for collaborative

entreprises, p.5, Linz, Australia).

Molina Espinosa J. M, 2003b. "Modèles et services pour

la coordination des sessions coopératives multi

applications: application à l'ingénierie systèmes

distribués". Thèse de doctorat en Informatique et

télécommunications, LAAS of CNRS, Toulouse.

Nielsen M., Plotkin G.D, Winskel G., 1981. "Petri Nets,

Event structures and Domains". Part I of Theoretical

Computer Science, Volume 13, N 1, pp. 85-108.

Orvalho J., Andrade T., Boavida F., 1999. "A conference

Service Based on the CORBA Event Service".

Proceedings of the 2nd Conference on

Telecommunications, Telecommunications Institute de

Portugal, Sesimbra, Portugal.

Rodriguez Perlta L.M., Villemur T., Drira K., Molina-

Espinosa J.M., 2002. "Managing dependencies in

dynamic collaborations using coordination diagrams".

Rapport LAAS N 02527, 6th International Conference

on Principles of Distributed systems, pp. 29-42,

Reims, France.

Roussel N., 1997. "Au dela du Madiaspace: Un Modèle

pour la collaboration médiatisée", In Actes des

neuvieme Journee francophones sur l'Interaction

Homme Machine (IHM'97), pp. 159-166, Futuroscope.

Tata S., 2002. "Policies for Cooperative Virtual Teams".

Proceedings of the 5th International Conference,

COORDINATION 2002, York, UK, LNCS 2315, pp.

340-347.

Wilde E., Freiburghaus P, Koller D., Platter B., 1996. "A

Group and session Management System for

Distributed Multimedia Applications, Multimedia

Telecommunications and Applications". Third

International COST 237 Workshop, Barcelona, Spain,

LNCS 1185, pp. 1-22.

Winskel G., 1987. "Event Structures". In Advances in

Petri nets, LNCS N. 255, pp. 325-392, Springer-

Verlag.

Winskel G, Nielsen M., 1992. "Models for concurrency".

The Handbook of Logic in Computer Science,

Technical Report DAIMI PB-429, Computer Science

Department, Aarhus University.

AN EVENT STRUCTURE BASED COORDINATION MODEL FOR COLLABORATIVE SESSIONS

143