A NETWORK COMPONENT ARCHITECTURE FOR

COLLABORATION IN MOBILE SETTINGS

Bernd Eßmann, Thorsten Hampel, Thomas Bopp

Heinz Nixdorf Institute

University of Paderborn

urstenallee 11, 31102 Paderborn

Keywords:

Mobile Computing, Spontaneous Collaboration, Distributed Knowledge Spaces, CSCW.

Abstract:

Today Computer Supported Cooperative Work (CSCW) is used in broad areas of human cooperation. With the

propagation of radio-based communication and ad hoc networking it may enter new areas of human cooper-

ation. One important aspect is the new quality in CSCW of being independent from special network-enabled

places. Another aspect is the more intuitive support of face-to-face cooperation utilizing personal mobile de-

vices. To open this ﬁeld of collaboration our approach featuring Distributed Cooperative Knowledge Spaces

speciﬁcally addresses conceptual issues pertaining to the transition from classical, server-centered to mobile,

distributed collaboration environments. With this concept we introduce persistent and personal knowledge

spaces as well as so-called temporary knowledge areas and groups. Our prototypical application for sponta-

neous collaboration implements this approach. We are able to draw here on many years of experience in the

development and testing of our concept of Cooperative Virtual Knowledge Spaces.

1 INTRODUCTION

In recent years, the CSCW world has created sev-

eral cooperation environments to support inter-human

work using network technology provided by the In-

ternet. We, for our part, have developed the sTeam

system, based on the concept of Cooperative Vir-

tual Knowledge Spaces. Knowledge Spaces allow

users to structure their knowledge spatially in virtual

rooms (also called ”areas”). sTeam is designed as

a client/server system with a central server provid-

ing the cooperation environment, which is accessible

from any device connected to the Internet. The idea is

that users should be able to cooperate independently

of speciﬁc locations such as conference rooms.

The need for collaboration environments that are

independent of speciﬁc locations is the reason for

our conviction that network-supported cooperation

will change in the near future. Ad hoc networks

are encouraging the emergence of this feature. Such

networks allow devices to be connected sponta-

neously without planning an infrastructure. This

makes network-based cooperation environments in-

dependent of network-connected locations.

One problem we face when using these new net-

work topologies is that data is often distributed across

several devices. For our cooperation environment,this

means that the Knowledge Spaces of the cooperation

partners are also distributed across their devices. The

original sTeam system ensured consistency of the un-

derlying data by managing it through one server. It

was not planned to store the data on multiple servers.

This means that areas of the Distributed Knowledge

Spaces can be connected across several devices.

An additional problem is that the availability of de-

vices in an ad hoc network is not guaranteed because

its structure is highly dynamic. This means that parts

of the Knowledge Spaces could also disappear when

hosted on a vanishing device.

One of our goals, then, is to expand our architecture

to cover these network requirements, allowing us to

create Knowledge Spaces across several devices with-

out much need for administration by the users. An-

other is to support both the classical, proven concepts

of CSCW and new concepts for spontaneous network-

ing. By doing so, we seek to transfer our well-proven

concept of Cooperative Virtual Knowledge Spaces to

such a network environment.

After consideration of related work, Section 3 of

this paper describes future network structures based

on MIP and Multihop Ad Hoc Networks (MANETs).

Such a new network structure will change the pos-

337

Eßmann B., Hampel T. and Bopp T. (2004).

A NETWORK COMPONENT ARCHITECTURE FOR COLLABORATION IN MOBILE SETTINGS.

In Proceedings of the Sixth International Conference on Enterprise Information Systems, pages 337-343

DOI: 10.5220/0002622503370343

 SciTePress

sible concepts for the desired cooperation environ-

ment. For this reason, we discuss the new concept

of Distributed Knowledge Spaces in Section 4. This

new concept extends Cooperative Virtual Knowl-

edge Spaces to include distributed environments like

MANETs without neglecting existing structured net-

works like the Internet. In addition, this concept is ca-

pable of combining both network topologies and en-

riching the concepts for classical infrastructures.

These two factors, the network structure and the

concepts, inﬂuence the design of the application for

the cooperation environment. We introduce and dis-

cuss this design in Section 5. We present the technolo-

gies used, JXTA, mDNS and sTeam, as well as initial

prototypes for evaluating the technology and our con-

cepts.

In the ﬁnal section, we discuss our experience so

far and look at the tasks ahead.

2 RELATED WORK

Using mobile devices in cooperation scenarios is an

attractive prospect. Our research focuses on develop-

ing a collaborative working environment using spon-

taneously connected devices with the option of ac-

cessing services in structured networks like the Inter-

net. Many projects explore the use of spontaneous

networks.

The ELAN project

aims to develop an infrastruc-

ture for e-learning in mobile ad hoc networks (Lauer

and Matthes, 2002). Unlike our approach featuring

distributed knowledge spaces for collaborative knowl-

edge construction in mobile settings, this project fo-

cuses on pure e-learning issues in these networks.

The project ConcertStudeo being conducted by the

Fraunhofer Integrated Publication and Information

Systems Institute’s (IPSI) Concert work group com-

bines an electronic blackboard with handheld devices

for enhancing face-to-face learning using interactive

learning tools (Wessner et al., 2003). This approach

is based on special tools (e.g. voting tools) for speciﬁc

learning settings. The cooperation concept focuses on

e-learning only; it does not use ad hoc networks for

the learning environment.

Another IPSI project, called MIRIAM, aims to set

up a reference installation for Mobile IP over IPv4

(MIPv4). Concrete installations based on MIPv4

are evaluated in an application scenario for scien-

tists (Berier 2001). The results of this project can

help us to select a suitable network for our research

The ELAN project is part of the German Research

Foundation’s (DFG) Priority Programme 1140 ”Basic Soft-

ware for Self-Organizing Infrastructures for Networked

Mobile Systems”.

needs. Although the project analyzed mobile net-

working, it did not address ad hoc networks or focus

on e-learning scenarios.

The MIT has a research project called Oxygen

addressing various aspects of pervasive computing.

Oxygen seeks to enable pervasive, human-centered

computing by combining speciﬁc user and system

technologies. The term ”ambient intelligence” is used

in this context (Dertouzos, 1999; Hanssens et al.,

2002). The goal is to build an intelligent ambi-

ent environment enriched with a collaboration and

knowledge-access subsystem. This project does not

exploit the potential of Cooperative Virtual Knowl-

edge Spaces for knowledge structuring and manage-

ment as we do.

The above projects focus on one or two aspects of

the research needed to build, as our approach seeks to,

an environment based on ad hoc established and struc-

tured networks allowing users to tailor their own Co-

operative Virtual Knowledge Spaces with a minimum

of administrative effort. What we need is a compre-

hensive concept of Cooperative Virtual Knowledge

Spaces as in classical sTeam, ported to mobile dis-

tributed environments. This concept must take into

account hardware issues, network technology and ap-

plication frameworks, enabling scaling from powerful

hardware down to ultra-portable devices like PDAs.

3 NETWORK TOPOLOGIES

SUPPORTING MOBILITY

An important factor in the design of the collaborative

work environment is the nature of the underlying net-

work structure. Put another way, the design of the

collaborative system inﬂuences the choice of network

structure. In our case, the idea is to use available net-

work technology, adapting it to match future network

structures and meet the needs of mobile learning en-

vironments. In our view, the network infrastructure

of the future must offer both maximum independence

from established infrastructures and the option of ac-

cessing classical networks like the Internet.

As cooperation occurs spontaneously, it should not

depend on existing network infrastructures. The net-

work infrastructure must be built from scratch, with

no need to conﬁgure the network devices manually.

Ad hoc networks meet these requirements. They set

up the network devices and establish routing schemes

appropriate to the network structure. Ad hoc networks

are able to maintain changing network infrastructures

on the ﬂy when devices move within the network or

disappear.

Although ad hoc networks would appear to be the

ideal solution for spontaneous collaboration, mobile

users still beneﬁt from connecting to classical net-

ICEIS 2004 - SOFTWARE AGENTS AND INTERNET COMPUTING

338

work infrastructures like the Internet. Many use-

ful services are only available in classical networks,

and remote cooperation over long distances can be

achieved through Internet connection. This is why

we wish to retain the option of accessing the classi-

cal sTeam server, when reachable.

By combining these technologies, three criteria are

met:

1. spontaneous networking without existing network

infrastructure

2. the use of existing network structures, if reachable

3. availability of the technology in near future, based

on already existing network structures and tech-

nologies

In a setting incorporating ad hoc and structured net-

works, reachable devices must be contactable through

both network topologies. A unique worldwide contact

address is needed because a device can move freely

within structured and ad hoc networks. Movement

within structured networks is called macro or inter-

domain mobility, movement within ad hoc networks

micro or intra-domain mobility.

The Internet Engineering Task Force (IETF) work

group’s Mobile IP (MIP) is a network technology pro-

viding macro mobility and reachability, irrespective

of the network point to which the device is connected

(Perkins, 1997; Perkins, 1998). MIP was originally

designed for IPv4 (MIPv4) and can be used for IPv6

(MIPv6) (Johnson et al., 2003).

While MIP supports macro mobility, it is not

suitable for micro mobility and lacks independence

from established networks. Mobile ad hoc networks

(MANETs) are therefore used for building networks

without the need to connect to established networks

(Perkins, 2001). If an MIP agent is in the range of the

MANET, devices able to connect to the agent can pro-

vide access to the structured network as a service to

the other devices in the MANET (Tseng et al., 2003).

In this case, a bridge between the two networks is es-

tablished. As soon as this bridge exists, all clients of

the ad hoc network are able to connect to their home

agent and are reachable from outside . Conversely, the

loss of the bridge does not affect the ad hoc network;

communication between the ad hoc-attached devices

is still possible (Campell et al., 2002).

In recent years, several routing protocols for

MANETs have emerged. They allow establishment

of the previously discussed spontaneous networks. So

far, however, none of them has been established as

a standard for ad hoc networking

. The only proto-

col for establishing ad hoc networks available in most

The Ad hoc On-Demand Distance Vector (AODV)

Routing is speciﬁed in an RFC (Perkins et al., 2003), but

is not available as standard protocol stack in current operat-

ing systems.

operating systems is the link-local protocol (Cheshire

et al., 2003)

. The main drawback of the link-local

protocol is that it is not a multi-hop protocol. This

means that it is not able to build larger ad hoc net-

works. Our approach is to use it for spontaneous net-

working in small groups, which meet at one location

and are within range of all other radio network de-

vices.

For bigger scenarios, it should be possible to es-

tablish multi-hop networks like the one mentioned in

(Perkins, 2001), where every device can be a router

for other devices, which would allow ad hoc networks

to be established over large areas. This would allow

connection to the Internet with mobile IP. Our short-

term goal is to create some test installations of this

network to evaluate them in everyday use.

In the described networks, client/server applica-

tions would make no sense because loss of the server

from the network would render all affected clients

useless. Unlike client/server structures, peer-to-peer

technology meets the needs of such a network topol-

ogy perfectly.

Knowing the characteristics of the network in-

frastructure for the learning environment, we must

integrate our current cooperative knowledge spaces

into this combination of structured networks and

MANETs. Our goal is to integrate the current con-

cepts of cooperative knowledge spaces into this inno-

vative network topology.

4 CHALLENGES OF

COOPERATIVE KNOWLEDGE

SPACES IN MOBILE

ENVIRONMENTS

The concept of Cooperative Virtual Knowledge

Spaces has been successfully applied for a number

of years. Our work aims to provide learners with

Distributed Knowledge Spaces based on innovative

network infrastructures for mobile devices. We be-

gin by evaluating the potential of handheld devices to

provide user interfaces for virtual knowledge spaces,

go on to address the task of building an environment

for mobile collaborative learning, and ﬁnally discuss

our concepts for distributed cooperative knowledge

spaces and spontaneous learning sessions.

Our goal is to facilitate state-of-the-art mobility-

supporting network technology, as described in Sec-

tion 3, and evaluate it in terms of our aims. Based on

these networks, we seek to build an application that

allows users to cooperate with a minimum of admin-

istrative effort.

The link-local protocol is part of MacOS X, Linux and

newer Windows versions

A NETWORK COMPONENT ARCHITECTURE FOR COLLABORATION IN MOBILE SETTINGS

339

Handheld devices are small enough to support

portability, often have enough power to compute the

software for these environments, and the wireless net-

work interfaces now frequently included allow con-

nectivity with other devices. To evaluate the spe-

cial characteristics of PDAs, we developed clients for

our sTeam client/server environment. While the ﬁrst

prototypes were HTML-based (Eßmann and Hampel,

2003), the newer generation is Java-based, to beneﬁt

from sTeam’s event mechanisms.

While the developed user interfaces meet the needs

of PDAs? small displays and pen-based input, they

are still classical clients in a client/server infrastruc-

ture. Parallel to this, we built the middleware for

ﬁnding other peers and communicating with them in

MANETs and structured networks. We describe this

middleware in Section 5 and it is the main focus of

this paper. Our application is thus designed as a peer-

to-peer architecture capable of running on mixed net-

work topologies .

The novel feature of using CSCW environments for

mixed network structures is not only the advantage

of independence from speciﬁc locations but also the

loss of guaranteed services within these spontaneous

networks. This loss of guarantees poses a number of

challenges in terms of adapting concepts from clas-

sical CSCW environments. In peer-to-peer topolo-

gies, all resources for needed services must be able

to be hosted on all clients. An important problem

are persistent knowledge areas of classical sTeam,

in which learners cooperate by communicating and

managing documents. We need a concept to distribute

the knowledge space over the mobile devices in such a

way that the required knowledge areas are accessible

to all users.

Spontaneous networks are by nature highly dy-

namic. The existence of a node in this network is

not guaranteed. The node containing the desired

knowledge area must not be a member of the net-

work or may suddenly disappear from it. Whole

parts of a network may also be separated from each

other. In (Feeney et al., 2001), this problem is called

network partitioning.

Since users can move around, the availability of a

certain knowledge area is not predictable. Given the

uncertain availability of knowledge areas on foreign

hosts, users may want to take interesting areas with

them. To do so, users need their own instances of

interesting knowledge areas to work with in case

they are disconnected from their co-workers. This

results in redundant storage of the objects involved.

If one user changes the content of an object within

a certain area, all other instances become outdated.

Problems may arise if two instances are combined

again when two peers reconnect with each other.

This is what (Feeney et al., 2001) calls network merg-

ing. If the instances of an area change, this poses the

Proxy Object

Persistent Area

Local Copy for Backup

Active Connection

Backup Connection (Read Only)

Synchronizing Connection (One Way)

Hosting Peer Remote Peer

Connected Peers

Disconnected Peers

Figure 1: Remote and hosting peer when connected and

when disconnected

challenge of synchronizing them between the affected

nodes.

In order to explore practical solutions to these prob-

lems, we began by adopting the pragmatic approach

of creating persistent and personal knowledge areas.

Our next step was to design temporary knowledge ar-

eas and temporary user groups.

4.1 Persistent and Personal

Knowledge Areas

In classical client/server environments, all users work

on the same instance of a knowledge area. This in-

stance remains persistent on the server until an au-

thorized user destroys it. These persistent knowledge

areas can be guaranteed because all users work on the

same objects. In ad hoc networks, there is no such

guarantee, not even for personal work areas. In Dis-

tributed Knowledge Spaces, we distinguish between

personal and persistent knowledge areas.

Personal knowledge areas are hosted only on the

learners’ personal device and are only accessible

to them. Unlike personal knowledge areas, persis-

tent knowledge areas are areas for cooperative work

hosted on precisely one node in the network (hosting

peer). All connected nodes (remote peers) may per-

form changes within this area through a proxy object

pointing to the original area. As shown in Figure 1,

nodes are also allowed to create a synchronized lo-

cal read-only instance in case they are disconnected.

When ofﬂine from the hosting peer, the remote peer

may read-access the local copy.

The problem of local copies becoming outdated

poses a challenge that it is hard to meet with tech-

nical solutions. Classical synchronization for mobile

clients as in PalmSync or CPISync is designed to syn-

chronize changing datasets between only two or three

ICEIS 2004 - SOFTWARE AGENTS AND INTERNET COMPUTING

340

Device n

Device 1

...

Temporary Knowledge Spaces Persistent Knowledge Spaces

...

Figure 2: Temporary knowledge areas may reference per-

sistent knowledge areas but not vice versa

hosts in one or the other direction (Trachtenberg et al.,

2002). Dynamic synchronization of a complete net-

work’s nodes is an even tougher task. Our concept

enables this problem to be evaded.

4.2 Temporary Knowledge Areas

and Groups

Another difference from classical learning environ-

ments is the fact that in mobile scenarios collabora-

tive work is often not planned with the goal of creat-

ing persistent knowledge spaces for future work but

to use knowledge spaces for spontaneous knowledge

construction. Consequently, the developed knowl-

edge space can be destroyed as soon as the session

is ﬁnished. It may be the case that one or more partic-

ipants in the session wish to keep some of the results

or even the whole area. Since this cannot be decided

when the session starts, it must be possible to trans-

fer the results of the session, or even parts of it, to

a persistent knowledge space. The same applies to

the spontaneously established groups because the de-

cision to hold a subsequent session may have been

reached after the group was established.

For this reason, our concept allows temporary

groups to be established with an appropriate knowl-

edge area. This mechanism is specially designed for

spontaneous face-to-face learning sessions, where all

participants meet at a certain location and wish to col-

laborate. The users only need their connected devices.

An important restriction on temporary knowledge

areas is the linking of objects. While in classical

sTeam systems any object may be referenced by a

link, links to temporary knowledge areas are prohib-

ited (see Figure 2. To reference a temporary knowl-

edge area, it must ﬁrst be transformed into a persistent

knowledge area.

Like the area, the temporary group exists as long as

at least one user is participating in the session. When

the last participant leaves, the group is automatically

destroyed. If the hosting device withdraws from the

network, the session still exists but the other partic-

ipants have only a read-only copy, as in the case of

persistent knowledge areas. These conventions ensure

that only one active instance of the area exists.

This concept is evaluated by implementing proto-

types for the cooperation environment, as described

in the following section. Evaluation results relating to

the concept of Cooperative Virtual Knowledge Spaces

can be found in (Hampel and Keil-Slawik, 2003).

5 AN ARCHITECTURE FOR

MOBILE COLLABORATION

The last two sections discussed the network environ-

ment for our application and described collaboration

techniques matching such network environments.

The devices the application has to run on may be

as heterogeneous as the network. They include any

network-connected device, ranging from cell phones

and handheld devices to full-featured PCs and servers.

This heterogeneous situation yields some important

criteria for our application:

Open Architecture The application must be portable

to almost any platform/device. Standardized proto-

cols and interfaces must therefore be used. Also, it

must scale to the capabilities of the hosting devices.

Automatic Conﬁguration The application must be

able to conﬁgure to meet the requirements of the

actual environments with a minimum of adminis-

trative effort. Also, it must reconﬁgure dynami-

cally for changing environments.

Spontaneous Networking The application must able

to establish spontaneous sTeam networks within

heterogeneous IP-based network environments.

This includes the option of sharing resources within

the network and connecting to classical sTeam

servers.

The goal of our development process is to build an

open, self-conﬁguring and well-scaling architecture.

Our architecture is based on our work on the classical

client/server sTeam system, which has been shown to

work well in common IP-based network infrastruc-

tures. Since every resource in sTeam is represented

by an steam-object, which can be both a user in the

system and a knowledge area or document, we speak

only of objects, regardless of the kind of object we are

talking about.

A node in a sTeam a peer-to-peer network may have

limited and possibly low resources. For this reason,

nodes must manage resources very carefully. Typi-

cal critical resources on mobile nodes are memory,

computing power (CPU), battery life and bandwidth

(Meyer, 1995).

One component in our architecture manages the re-

sources of a node. We call this component Resource

Handler. The Resource Handler warns about low

resources and is conﬁgured with policies on how to

A NETWORK COMPONENT ARCHITECTURE FOR COLLABORATION IN MOBILE SETTINGS

341

Peer

Battery

CPU

Other

Resources

Memory

Resource Handler

JXTA COAL

Other

Services

mDNS

sTeam Peer-to-Peer

IP-based Services

sTeam Server

Network Application Hardware

Figure 3: The Resource Handler controls the network com-

ponents the peer may offer to the network according to the

available hardware resources

take care of resource conﬂicts. An example policy is

used for peers with low battery power. The Resource

Handler invokes the migration of shared objects from

the low-energy peer to another peer with better bat-

tery resources. This allows the drained peer to save

battery power by reducing its network activity while

other remote peers can still work on the shared ob-

jects. Even if the node disappears from the network

because of its low battery power, it does not affect

the remotely working peers. Similar policies exist for

low memory, CPU performance, bandwidth and any

other resource monitored by the Resource Handler.

The Resource Handler balances available resources

between nodes by communicating with its counter-

parts in other sTeam peers via the network.

Another way to save a node?s resources is to ac-

tivate/deactivate several services to the network, as

shown in Figure 3. Thus, a node on a low-featured

device like a mobile phone would merely access other

remote resources but offer no services to other nodes.

Better equipped devices can even offer conventional

services to access a knowledge area (e.g. ftp access).

We will discuss this option later on.

To build our application, we opted for the JXTA

peer-to-peer framework, which implements a over-

lay peer-to-peer network that allows communication

between any network-connected device, from cell

phones and handheld devices to personal computers

and servers. Figure ?? shows a small JXTA network.

By using JXTA applications are able to communicate

on a peer-to-peer basis without knowing the underly-

ing network structure (Gong, 2001).

An important requirement for our application is the

ability to connect automatically to its counterparts to

form a spontaneous application network. JXTA shows

this behavior even in a hybrid architecture of struc-

tured and ad hoc networks.

While JXTA offers services to other JXTA nodes,

other IP-based applications are unable to access the

resources on the sTeam peer. To be as cooperative as

possible with other parts of the network, our applica-

tion is able to offer services to classical clients, e.g.

web browsers or ftp clients. These additional services

can be provided depending on the capabilities of the

node. Furthermore, the sTeam peer is able to use stan-

dard IP-based services for its purposes. This leads to

the problem of ﬁnding services in ad hoc networks

where no ﬁxed namespaces exist and where access to

DNS servers cannot be guaranteed.

For publishing and ﬁnding such services, we use

Multicast DNS (mDNS). mDNS broadcasts informa-

tion about provided services to the network. This in-

formation can be read by any mDNS-capable applica-

tion to determine which IP address and port to connect

to for which service. mDNS is part of the Zeroconf

speciﬁcation carried out by IETF’s Zeroconf Working

Group and is described in (Cheshire and Krochmal,

2003).

The second optional network protocol is the COAL

protocol. This is the proprietary standard sTeam pro-

tocol for connecting to classical sTeam servers and

clients. Thus, the peer can communicate with sTeam

servers and provide objects to classical sTeam clients.

The use of optional services allow peers to scale

their provided services to suit their resource capabil-

ities. In addition to the above-mentioned resources,

the Resource Handler also controls the provided net-

work services according to available bandwidth, com-

puting power, memory, etc. This leads to high scala-

bility. The application can scale both on the JXTA

part by scaling from Minimal Peer up to Rendezvous

or Relay Peer (Gong, 2001) and on the sTeam part

by scaling from a pure remote peer, only working on

knowledge areas hosted on other nodes, up to a net-

work server, offering access to its knowledge areas via

http, ftp, COAL, etc. This concept enables the applica-

tion to adapt to devices like cell phones, full-featured

laptops or even workstations.

As part of our development process, we built a pro-

totype of an sTeam peer node using the JXTA tech-

nology. The prototype is quite a simple peer, which

is able to send messages containing XML structures,

steam-objects or Java objects from one peer to an-

other. This is done without any administrative effort

on the part of the users. The peers ﬁnd each other

autonomously without any conﬁguration. This proto-

type shows that it is feasible to build an sTeam peer-

to-peer network with the basic required communica-

tion mechanisms using the JXTA framework.

Parallel to this, we built a Java-based framework

for implementing mDNS-capable Java applications. It

is based on the Multicast DNS implementation jRen-

ICEIS 2004 - SOFTWARE AGENTS AND INTERNET COMPUTING

342

dezvous of Strangeberry Inc.. Our framework pro-

vides a Java interface class, which must be imple-

mented by an application to provide an mDNS service.

As with the JXTA prototype, instances of the chat ap-

plication are able to ﬁnd each other and establish con-

nections without any conﬁguration by the users. With

our mDNS interface and the JXTA prototype, we now

have the basic components for our sTeam peer-to-peer

application. We hope to be able to present the results

of our ongoing work at the conference.

6 CONCLUSION

The described prototype for collaboration in mobile

ad hoc networks constitutes a further step toward

network-supported, location-independent collabora-

tion. Our concept enables the users of mobile col-

laboration environments to cooperate via their mo-

bile devices without the need for any existing network

infrastructure, and at the same time allows them to

use classical services when available. By doing so,

we avoid the trade-off involved in using spontaneous

peer-to-peer networks or established client/server so-

lutions.

Both concepts are consolidated in our approach.

Our prototype demonstrated that it is possible to

transfer our well-proven concept of Cooperative Vir-

tual Knowledge Spaces from common network infras-

tructures to the innovative hybrid network structures

described in Section 3. To do so, we are adapting

Cooperative Virtual Knowledge Spaces to dynamic

network structures. Our goal is to develop a con-

cept for Distributed Knowledge Spaces. Distributed

Knowledge Spaces allow Knowledge Spaces to be dis-

tributed over multiple dynamic connected nodes in

spontaneous and classical networks.

Computer-supported cooperation can thus accom-

pany users through the sort of situations encountered

in their daily lives. This is our primary research goal.

REFERENCES

Campell, A. T., Gomez, J., Kim, S., Wan, C.-Y., Turanayi,

Z. R., and Valko, A. G. (2002). Comparison of ip

micromobility protocols. IEEE Wireless Communica-

tions, 9(1):72–82.

Cheshire, S., Aboba, B., and Guttman, E. (2003). Dy-

namic conﬁguration of link-local ipv4 addresses.

Http://todo/draft-ietf-zeroconf-ipv4-linklocal-08.txt

(june 03 2003), Zeroconf/DNSEXT Working Group

(IETF).

Cheshire, S. and Krochmal, M. (2003). Performing dns

queries via ip multicast. http:/todo/draft-cheshire-

dnsext-multicastdns-02.txt (june 20 2003), Zero-

conf/DNSEXT Working Group (IETF).

Dertouzos, M. (1999). The oxygen project. Scientiﬁc Amer-

ican, 282(3):52–63.

Eßmann, B. and Hampel, T. (2003). Human computer in-

teraction and cooperative learning in mobile environ-

ments. In Human Centred Computing Cognitive, So-

cial and Ergonomic Aspects. Proceedings of the HCI

International 2003, pages 694 – 698.

Feeney, L. M., Ahlgren, B., and Westerlund, A. (2001).

Spontaneous networking: An application-oriented ap-

proach to ad hoc networking. IEEE Communications

Magazin, 39(6):176–181.

Gong, L. (2001). Jxta: A network programming environ-

ment. IEEE Internet Computing, 5:88–95.

Hampel, T. and Keil-Slawik, R. (2003). Experience with

teaching and learning in cooperative knowledge areas.

In Proceedings of the Tracks Education, Global Com-

munity, Industrial, Practice and Experience,Web Ser-

vices, pages 125–132. WWW2003. The Twelth Inter-

national World Wide Web Conference.

Hanssens, N., Kulkarni, A., Tuchinda, R., and Horton, T.

(2002). Building agent-based intelligent workspaces.

In Proceedings of the ABA Conference, page To Ap-

pear.

Johnson, D., Perkins, C. E., and Arkko, J. (2003). Mobil-

ity support in ipv6. Internet draft, draft-ietf-mobileip-

ipv6-24.txt, june 30, 2003, IETF.

Lauer, M. and Matthes, M. (2002). Elan: An e-learning

infrastructure for ad-hoc networks. In Proceedings of

the eighth ACM International Conference on Mobile

Computing and Networking, volume 7, pages 53–55.

Meyer, A. (1995). Pen computing. SIGCHI Bulletin,

27(3):46–90.

Perkins, C., Belding-Royer, E., and Das, S. (2003). Ad hoc

on-demand distance vector (aodv) routing (rfc 3561)

(july 2003). Rfc, IETF Network Working Group.

Perkins, C. E. (1997). Mobile IP: Design Principles and

Practices. Addison-Wesley.

Perkins, C. E. (1998). Mobile networking in the internet.

Mobile Networks and Applications, 3(4):319–334.

Perkins, C. E., editor (2001). Ad Hoc Networking. Addison

Wesley, Boston, USA.

Trachtenberg, A., Starobinski, D., and Agarwal, S. (2002).

Fast pda synchronization using characteristic polyno-

mial interpolation. In Proceedings of the INFOCOM

2002. 21th Annual Joint Conference of the IEEE Com-

puter and Communications Societies, pages 1510–

1519.

Tseng, Y.-C., Shen, C.-C., and Chen, W.-T. (2003). Integrat-

ing mobile ip with ad hoc networks. IEEE Computer,

pages 48–55.

Wessner, M., Dawabi, P., and Fern

andez, A. (2003). Sup-

porting face-to-face learning with handheld devices.

In Designing for Change in Networked Learning En-

vironments, Proceedings of the International Confer-

ence on Computer Support for Collaborative Learn-

ing 2003, pages 487–491.

A NETWORK COMPONENT ARCHITECTURE FOR COLLABORATION IN MOBILE SETTINGS

343