=Paper=
{{Paper
|id=Vol-505/paper-4
|storemode=property
|title=Specification and Analysis of Requirements Negotiation Strategy in Software Ecosystems
|pdfUrl=https://ceur-ws.org/Vol-505/iwseco09-4Fricker.pdf
|volume=Vol-505
|dblpUrl=https://dblp.org/rec/conf/icsr/Fricker09
}}
==Specification and Analysis of Requirements Negotiation Strategy in Software Ecosystems==
https://ceur-ws.org/Vol-505/iwseco09-4Fricker.pdf
19 Proceedings of the first International Workshop on Software Ecosystems 2009
Specification and Analysis of Requirements Negotiation
Strategy in Software Ecosystems
Samuel Fricker1,2,
1
University of Zurich, Binzmuehlestrasse 14, 8050 Zurich, Switzerland
fricker@ifi.uzh.ch
2
FUCHS-INFORMATIK AG, Rankweg 10, 4410 Liestal, Switzerland
samuel.fricker@fuchs-informatik.ch
Abstract. The development of software products and systems generally
requires collaboration of many individuals, groups, and organizations that form
an ecosystem of interdependent stakeholders. The way the interests and
expectations of such stakeholders are communicated is critical for whether they
are heard, hence whether the stakeholders are successful in influencing future
solutions to meet their needs. This paper proposes a model based on negotiation
and network theory for analyzing and designing flow of requirements through a
software ecosystem. The approach supports requirements engineering process
engineers and managers in taking strategic decisions for resolving
communication bottlenecks, increasing overall requirements engineering
productivity, and consciously assigning power to stakeholders.
Keywords: Requirements Engineering, Process Improvement, Negotiation,
Network Management.
1 Introduction
Large-scale organizations need to consider the interplay of a considerable number of
stakeholders for defining requirements of their commercial and technical products and
product platforms [1]. Different specifications are used to negotiate and document
agreements that align interests and expectations of these stakeholders. Examples are
marketing requirements specifications to define the product-related offering by
product management towards key account managers and customers, use case
specifications to align product management and product users, technical
specifications to align development management and product management, and
system specifications to align team leaders and development management [1].
The ecosystem considered in this article is the stakeholder network typically
encountered in large software product organizations and their markets [2].
Requirements communication networks (RCN) are proposed to describe how a given
ecosystem is structured in terms of interdependent stakeholders that negotiate
requirements for aligning needs with solutions, solution components, and solution
portfolios. Such a network definition can assist process improvement by enabling
identification and resolution of communication-related problems and by prescribing
desired network structure and stakeholder behavior [3].
Cite as: Fricker, S. (2009). Specification and Analysis of Requirements Negotiation Strategy in Software Ecosystems. In
Proceedings of the First Workshop on Software Ecosystems (pp. 19-33), CEUR-WS.
20 Proceedings of the first International Workshop on Software Ecosystems 2009
Acceptance of an ultimately developed product requires agreement among
stakeholders regarding the desired capabilities and impacts of the product. One
approach to reach such agreement is the use of integrative negotiation techniques [4].
The constitution of the negotiating stakeholders influences how agreements are
reached by providing a basis to choose communication tactics and techniques [5].
Current approaches for describing and analyzing stakeholder networks do not allow
differentiating such stakeholder constitutions. Stakeholder network modeling
languages employed for analyzing requirement flows and stakeholder dependencies
such as SSN [6], FLOW [3], i* [7], and E3Value [8] assume that stakeholders behave
like single negotiating parties and do not pay attention to the various types of multi-
party stakeholder groups [5]. Other stakeholder modeling approaches such as the
Onion model [9] and the VORD model [10] ignore relationships between
stakeholders, which are important to define communication channels.
This paper proposes a process analysis and improvement approach that is based on
modeling the requirements communication network of a software ecosystem.
Strategic decisions regarding network design are made based on computer network
theory. Tactical and methodical advice is provided to stakeholders based on the body
of knowledge of negotiation. The paper extends previous work [5] by refining the
modeling language, by introducing requirements communication strategy, and by
describing the application of the approach in a real-world exemplar.
The approach can be used to support software development governance [11] by
aiding managers and development process engineers in diagnosing problems related
to requirements communication within an ecosystem and in specifying desired
collaboration. The method selection and strategy evaluation framework helps to take
conscious decisions regarding communication structure and processes and allows to
record and access experience regarding such decisions.
The paper is structured as follows. Section 2 describes the background of the
presented work and motivates the requirements communication network approach.
Section 3 introduces necessary concepts and the modeling language to understand and
describe requirements communication networks. Section 4 describes the application
of the approach for process improvement. Section 5 summarizes and concludes.
2 Background and Motivation
Requirements communication is challenging, in particular when people and
organizations need to collaborate over dispersed locations. Some of the problems we
encountered were related to requirements communication tactics and methodology.
These problems concerned stakeholders that needed to communicate with each other
to achieve an agreed understanding of requirements and solutions [12]. Other
problems were related to a more strategic aspect of requirements communication, the
alignment of interests and expectations that is needed to prepare a company and its
markets to accepting a new software product, system, or service.
One of these strategic problems concerned a global Fortune500 company that
serves markets all over the world with software-intensive systems and has
development centers located in three continents. The organization’s product
21 Proceedings of the first International Workshop on Software Ecosystems 2009
management unit was reflecting how it best would address the elicitation of customer
needs and market trends to decide which markets it wanted to address with new
products and which customer needs it wanted to satisfy with new product features.
Figure 1 shows the structure of the organization and its markets.
Figure 1: Requirements communication network (notation: Table 1).
The concerned company was in a competitive environment, where markets with
differing needs are served by competing suppliers. For each market, a local product
manager was appointed to represent customer needs, competitive offerings, and
emerging trends. A senior product manager was responsible for integrating this
information into roadmaps and release plans of the products. The therein contained
themes and requirements were used as a basis to steer product development.
A key challenge was establishing the communication between the local product
managers and the senior product manager. The company experimented with different
technologies to facilitate information exchange between these parties, with limited
success. Requirement databases were not filled with information, e-mail was used in
an inconsistent manner, and travelling was expensive, hence done only rarely.
Process development in the described and in other organizations suffered from a
lack of approaches for modeling stakeholder networks and for supporting
management and process engineers with advice regarding design of requirements
communication. Process design was ad-hoc and method selection naïve. The missing
body of knowledge left the practitioners debating opinions instead of properly
analyzing problems, for which understood solutions can be devised and piloted.
3 Requirements Communication Networks
Requirements communication networks (RCN) enable analysis and design of
requirements communication among stakeholders of software products. The here
described approach provides a modeling language for specifying requirements
communication and a framework for evaluating communication strategy and selecting
tactics and methodology. The approach allows understanding how the interests and
expectations of given stakeholders can be communicated. The approach can be
22 Proceedings of the first International Workshop on Software Ecosystems 2009
integrated into process improvement, where it enables proactive and conscious
experience-based design of requirements communication.
3.1 Specification of a Requirements Communication Network
A requirements communication network (RCN) describes stakeholders and traces of
communicated requirements. A RCN describes the structure of a software ecosystem
in terms of actors, groups of actors, and negotiation paths between these groups of
actors. The presented approach takes a requirements negotiation perspective on
requirements communication by assuming that the requirements traces correspond to
agreements between stakeholders, respectively stakeholder groups, that are
documented in the form of specifications [5].
Figure 1 has introduced an example of such a RCN. Considered stakeholders are a
set of markets, the markets A and B, that consist of unlabeled customers. The markets
are served by a set of competing suppliers, the companies 1 and 2. Company 1 is
further refined into two competing local product managers, a senior product manager,
and a development organization. Figure 2 shows a corresponding organization chart.
The network shown in Figure 1 further introduced traces of communicated
requirements. Each arrow corresponds to an eventually reached agreement between
communicating stakeholders. An arrow points to the stakeholder that benefits from
assets of the stakeholder the arrow points from. An arrow between a market and a
local product manager corresponds to a specific product offering, the arrow between
the markets and company 2 to a competitive product offering, the arrow between the
local product managers and the global product manger to a market requirements
specification, and the arrow between the senior product manager and the development
organization to a technical requirements specification.
Stakeholders
Markets Suppliers
(differentiated) (differentiated)
Market A Market B
Company 1 Company2
(homogeneous) (homogeneous)
‐ Senior Product Product
(differentiated) Manager Developemnt
Local Product Local Product
Manager Manager
Figure 2: Stakeholder structure from Figure 1 shown as an organization chart
(refinements of markets A and B not drawn).
23 Proceedings of the first International Workshop on Software Ecosystems 2009
The RCN modeling approach uses the syntax described in Table 1. A single party
is a person or group that has one set of aspirations and one voice at the negotiation
table. No internal fragmentation exists: there is neither intrapersonal conflict of the
person nor interpersonal conflict in the cohesive group of people. A multi-party is a
group whose constitution influences the group’s negotiation behavior. The multi-party
group consists of single parties or again other groups that have individual voices at
the negotiation table. Agreements made by a multi-party at the primary negotiation
table should be ratified [5].
Table 1: Requirements communication network modeling syntax.
Stakeholders homogeneous collaborating differentiated
single-party
multi-party multi-party multi-party
2 2
Connectivity (assu- not 1 3 1 3
ming 5 members) constrained
5 4 5 4
agreement between customer and supplier roles
Relationships
(nesting) multi-party constitution
The constitution of a multi-party has an effect on the connectivity between the
party’s members. Members of a homogeneous multi-party have the same aspirations,
but individual voices, hence are not connected with each other. Members of a
collaborating multi-party have different aspirations but seek an agreement, hence are
fully connected. Members of a differentiated multi-party have different aspirations
and are competing with each other, hence again are not connected with each other.
Requirements are communicated across the stakeholder network along the
customer ← supplier relationships. Each such relationship represents an agreement of
interests and intentions between the two related parties that is established through
requirements negotiation [5]. In the case of homogeneous multi-parties, ratification of
an agreement needs to be done with every group member. With a collaborating multi-
party, ratification is part of finding a group-internal agreement. With a differentiated
multi-party, ratification involves establishing one agreement per member because
each member pursues different interests.
A globally accepted conceptualization of a RCN exists only if that one has been
specified and standardized. This can be achieved with process development. Non-
standardized networks live in the eyes of the stakeholders that engage in requirements
communication. These networks evolve when stakeholders change the peers they are
collaborating with and autonomously build ad-hoc groups and communication
channels. To provide an accepted and consistent view of requirements
communication, a network should be specified pragmatically by knowing the purpose
of the specification and by including just those elements that are needed for achieving
that purpose.
24 Proceedings of the first International Workshop on Software Ecosystems 2009
3.2 Advice for Requirements Communication
The network model provides value if it is combined with advice for how requirements
communication should be performed to achieve win-win agreements between
communicating stakeholders. A method and tactic selection framework has been
presented earlier [5] and is extended here with an approach for evaluating strategic
options.
Method selection is concerned with requirements communication from a bird’s-eye
perspective between a pair of communicating stakeholders [5]. Depending on the
constitution of the communicating parties, different methods are adequate. For
example a single-party supplier that communicates with a single customer is well
advised to employ Handshaking [12], with homogeneous customers market-driven
requirements engineering [2], with differentiated customers domain requirements
engineering [13], and with collaborating customers viewpoint-oriented requirements
development [10].
Tactic selection is concerned with requirements communication from an egocentric
perspective, where a stakeholder communicates with his peers and with his
negotiation partner [5]. Depending on whether the stakeholder is alone or forms a
multi-party with peers and depending on the constitution of his communication
partner, he selects different negotiation tactics to reach a win-win agreement. For
example, a single party uses a constituent tactic if he is communicating to another
single party or to a homogeneous multi-party, selects the member he prefers to agree
with if he communicates to a differentiated multi-party, and builds a coalition if he
communicates with a collaborating multi-party. These tactics can be studied in
standard textbooks on negotiation [14] and refine the generic approach to win-win
negotiations in the software domain [4].
The here introduced strategy evaluation considers the flow of requirements
between stakeholders that are not necessarily communicating directly with each other.
This level is concerned with the properties of end-to-end networks that are affected by
the network structure and the properties of relevant nodes that represent stakeholders
and links that represent communication channels, respectively agreements. Models for
describing such relationships are studied in the area of mesh communication networks
[15, 16]. In contrast to methods and tactics that are selected, strategy definition is a
design process, where different options are evaluated and validated. Section 3.3
elaborates the strategy level in more detail.
3.3 Requirements Communication Network Design
Requirements communication networks (RCN) share many characteristics of mesh
networks. A mesh network [15] is used to pass information and consists of nodes and
communication channels. Mesh networks can be structured in a predetermined
manner or are self-forming and self-organizing and can evolve over time. The social
nature of a RCN implies that strategic concerns related to shared decision-making,
hence negotiation, need to be considered.
Node and link properties and the network connectivity influence the properties of
the RCN. The better these relationships are understood the easier it is to design a
25 Proceedings of the first International Workshop on Software Ecosystems 2009
RCN that functions in a desired manner. Node properties of interest include intrinsic
node power and available node capacity. The here considered link property is link
efficiency. Structural network properties include extrinsic node power and link
dependency. Network properties that are affected by the former properties include
network capacity, load, latency, and reliability. Figure 3 summarizes.
Figure 3: Node and link properties and network structure influence end-to-end
network properties.
Node and Link Properties. Strategy-relevant factors related to given nodes and links
in the RCN include intrinsic node power, available node capacity, and link efficiency.
These factors affect communication reliability and latency of requirements that are
routed through the concerned nodes and links.
Intrinsic Node Power: Intrinsic power, also called referent power, expresses the status
of a stakeholder towards other stakeholders [14]. Such power is derived from respect
or admiration of the concerned person, group, or organization. Personality, integrity,
interpersonal style, or religious status such as membership of a high Indian caste lead
to such power. Referent power can be reinforced by appealing to common
experiences, common past, common fate, or group membership and is used to
persuade or dominate another party, hence to impose one’s own interests.
Actions to increase intrinsic node power include education, promotion, and better
integration of a stakeholder into the communication network. Intrinsic power can only
be indirectly reduced by isolating the stakeholder, by providing alternative
communication paths, and by reducing the stakeholder’s available capacity.
Available Node Capacity: Each node has limited capacity for handling given
requirements. Limitations are due to cognitive limits of people [17] and are affected
by work load devoted to activities other than the handling of the requirements under
consideration. For example, the upper level of complexity that an organization can
handle at a given moment in time has been suggested to lay between 1’000 and
10’000 requirements [18].
Actions to increase available capacity of a stakeholder include delegating the
handling of non-relevant requirements to other stakeholder, supporting the
26 Proceedings of the first International Workshop on Software Ecosystems 2009
stakeholder with assistants, and introducing effective requirements management tools.
As experience of a stakeholder grows over time, the efficiency of the stakeholder will
also grow. Actions to decrease available capacity of a stakeholders include reducing
supporting staff and assigning additional responsibilities other than the handling of
the requirements under consideration.
Link Efficiency: A link has limited capacity for transmitting given requirements. This
influences the time needed for communicating requirements from one stakeholder to
another stakeholder. Link efficiency is influenced by factors such as geographic
distance, knowledge of the communication partner, and trust.
Geographical distance reduces communication efficiency and effectiveness [19].
Problems caused by geographical distance include misunderstandings due to cultural
differences, low quality of communication channels, challenging knowledge
management, and time differences. These problems lead to inappropriate participation
of stakeholders in the communication process, lead to low awareness of local work
context, inhibit informal communication, and ultimately cause delay and
misunderstandings.
Trust between communication partners enables collaboration, and mistrust inhibits
it [14]. Mistrusting people are defensive, withhold information, and search for hidden,
deceptive meanings of information. Such behavior undermines the negotiation
process and stalls requirements communication.
The efficiency of communication between given stakeholders evolves over time
[12]. The longer parties collaborate, the more they learn about their background and
interests. This knowledge allows them to increase communication efficiency by
communicating more pragmatically and by better responding to gaps in the partner’s
requirements understanding [20]. Hence, the time needed to communicate
requirements successfully can be reduced over time.
Problems related to geographical distance can be mitigated by collocating
collaborating people. Communication techniques that minimize the need for physical
contact reduce the need for collocation and help saving travelling time and cost.
Trust can be built by acting in a cooperative manner and by believing that the
communication partner is committed to finding a joint solution [14]. Performing face-
to-face negotiation, rather than negotiations over distance, and sharing information
regarding the negotiation themes, and transparent fair acting facilitates trust building.
Trust, if broken, can be repaired by sincerely expressing apology and by taking
personal responsibility for the breach. Trust, however, can only be repaired if the
breach is an isolated event and if risk of deception is effectively mitigated.
Network Structure. Strategy-relevant factors related to the structure of the RCN
include extrinsic node power and link dependency. These factors affect routing
possibilities of the communicated requirements and define the degree to which given
node and link properties influence the overall requirements communication. Figure 4
shows special network topologies.
27 Proceedings of the first International Workshop on Software Ecosystems 2009
Figure 4: Network Topologies. The leaving and entering arrows indicate end points.
The role of special nodes are named according to negotiation literature [14].
Extrinsic Node Power: Extrinsic node power expresses the dependency on a given
node for successful requirements communication based on its connectivity to other
nodes in a network. Centrality and criticality characterize such power of a node [14].
The more connections the node has to other nodes, the more central it is to the
network. Centrality characterizes the influence a stakeholder can exercise to impose
its interests on other stakeholders and to route requirements. Centrality can be
achieved through controlling a large number of customer ← supplier relationships or
through a large number of memberships in multi-parties. Node 3 in the star topology
and all nodes in the full connected mesh have high centrality.
The more possible paths in a given network pass through a given node, the more
critical this nodes is for requirements communication. Criticality influences the
likelihood that the considered node participates in end-to-end communication of
requirements. Node 3 in the star topology and nodes 2, 3, and 4 in the line topology
have high criticality.
Actions to increase the extrinsic power of a given node include increasing its cross-
linking to other nodes, increasing its integration into groups, and excluding nodes that
can provide alternative communication paths from the network. Actions to reduce the
extrinsic power of a given node include isolation of the node and establishing
alternative communication paths that avoid that node. Such actions are particularly
important for successful requirements communication when trust in the
cooperativeness of that node is missing.
Link Dependency: Link dependency expresses the criticality of a given link for
successful requirements communication. Alternative routes help reducing the
dependency on a given link, hence increase the robustness of communication. The
availability of many redundant links, however, risks to introduce inconsistencies and
conflicts because the interests of a larger number of stakeholders have to be
considered in the end-to-end communication. The links 1-3 and 3-5 in the star
topology and all links in the line topology have high criticality. The redundancy of the
communication paths in the fully connected mesh implies that none of the mesh’s
links is critical.
Actions to reduce link dependency include the provision of new alternative
efficient communication paths and isolation nodes that communicate through the
concerned link. Such actions are particularly important for successful requirements
communication when using a link is costly, i.e. the link is inefficient, or when the
28 Proceedings of the first International Workshop on Software Ecosystems 2009
nodes that communicate over the link stand in conflict. Actions that can be taken to
increase link dependency include isolation of nodes that allow avoiding
communication through the concerned link.
End-to-End Network Properties. Concerns related to end-to-end network properties
that are of strategic relevance include network throughput, load, delay, and reliability.
These factors are quality of service aspects that affect the likelihood of whether a
given interest of the requirements source is considered by a solution deliverer and the
time it takes to have considered the interest.
Network Capacity: Network capacity is the number of requirements that can be
handled at a given moment. This property refers to the theoretical limit of traffic a
network can handle. If network load is below this limit, the network is underutilized
and more end-points should be connected. If network load is above this limit, the
network is in a state of congestion, where the performance drops drastically.
Network capacity needs to be traded off with end-to-end communication delay,
node capacity, and network cohesion. Each node can communicate only with a
limited number of partners and handle a limited number of distinct requirements. If
nodes have spare capacity, the network is underutilized. If the capacity of nodes is
exceeded, communication partners and requirements will be discarded or postponed
to a later moment where the communication with at least one active partners is ended.
Each communication channel, however, introduces delay. The more nodes a message
passes, the more time the message needs to traverse the network.
Actions to increase network capacity are concerned with restructuring the network.
If the amount of input cannot be handled and synergies between requirements are
more critical than network delay, additional intermediate nodes should be introduced.
If network delay is more critical than synergies between requirements, the network
should be split. Figure 5 illustrates these two actions for increasing network capacity.
Actions to decrease network capacity include removing intermediate nodes and
joining separate sub-networks. These actions should be taken if the network load falls
significantly below the network capacity.
Figure 5: Too many inputs (left) can be handled by chaining intermediate nodes
(middle) or reducing network cohesion (right). Assumed in the illustration is an upper
limit of three connections that can be handled concurrently by a node.
Network Load: Network load refers to the intensity of traffic of a network at a given
moment in time. Network load affects network delay and reliability. Network
congestion occurs when a link or node is carrying so much data. This results in a
deterioration of quality of service and effects delay, loss, or blocking of new
connections and requirements [16]. For reliable functioning of the network, it is
important to keep the network in a non-congested state.
29 Proceedings of the first International Workshop on Software Ecosystems 2009
Network load problems can be due to network design or due to problematic
implementation of requirements communication tactics and methodology. Bottlenecks
such as node 3 in the star topology of Figure 4 represent network design problems.
Communication inefficiencies can also result from inadequate requirements
communication tactics and methodology and from nodes with low power or with
uncooperative behavior.
Network load can be managed by controlling node and link utilization [16]. Load
balancing and congestion control can be used to distribute requirements
communication load over time and across the stakeholder network in a controlled
manner. Techniques include controlled forwarding and scheduling of development
requests and dedicating network resources to specific development themes or
products. They are applied to prevent congestion collapse, fair availability of
resources and services, and optimization of throughput, delay, and reliability.
Network Latency: One-way network latency is the time taken for a requirement
communicated at one end to be received at the other end. This property is sometimes
called lag or delay. Two-way network latency is the time taken for a requirement
communicated at one end to be answered by the other end. Two-way latency is
sometimes called cycle time.
One-way latency is the effect of aligning the interests of a chain of nodes link by
link by assuring agreement of adjacent nodes. Two-way latency is the effect of fully
aligning the chain of nodes by assuring agreement between any pair of nodes. The
more nodes need to be transferred for an end-to-end communication, the longer the
alignment of the chain takes. For example, the line topology of Figure 4 will take
longer time to align than the star or mesh topology. Full alignment is harder to
achieve and more effort-intensive than link-by-link alignment.
Network latency can be managed by adding or removing nodes needed to traverse
the network and by controlling link efficiency and available node capacity.
Network Reliability: Network reliability is the probability for a requirement
communicated at one end to be received at the other end. Network reliability is a
consequence of the time allowed for a requirement to traverse the network and the
capacity for processing and remembering requirements of the path’s nodes. Network
reliability may be different for requirements of different criticality.
Network reliability is influenced by the redundancy and reliability of
communication paths and the criticality and cooperation of the nodes on the path.
Nodes with maximal extrinsic power represent single points of failure whose loss or
non-cooperation leads to failed requirements communication. Examples are node 3 in
the star topology and nodes 2, 3, and 4 of the line topology in Figure 4. Links that a
network is highly dependent on represent other single points of failure. Examples are
the links 1-3 and 3-5 in the star topology and all links in the line topology in Figure 4.
The lower the efficiency of such a critical link is, the more problematic the alignment
of the communicating nodes is.
Network reliability can be managed by adjusting the network topology and by
controlling extrinsic node power and link dependency. The more redundant the
communication channels in a network are the more reliable the network is, but at the
expense of increased effort for maintaining the network. Alternatively, available
capacity, cooperation of relevant nodes, and link efficiency can be adjusted.
30 Proceedings of the first International Workshop on Software Ecosystems 2009
4 Process Improvement – an Exemplar
Process improvement in a requirements communication network (RCN) follows
roughly Plan-Do-Check-Act cycles [21] with iteratively performed diagnosis,
communication design, validation, and roll-out phases. This section shows how the
described network modeling and evaluation approach is applied in such process
improvement.
Diagnosis aims at understanding network structure and problems related to
requirements communication. Figure 1 and the description of the communication
challenge in section 2 are typical work results from such analysis. These results
document the basis for planning process changes and act as a reference to evaluate
impact that is eventually achieved with these changes.
Communication design aims at planning improvements in requirements
communication by identifying and prioritizing changes on the methodical, tactical,
and strategic levels. In the described case, the senior product manager should
implement domain requirements engineering [13] as a method to understand
commonalities and differences of the needs and expectations the local product
managers represent [5]. On the tactical level, she should prioritize and select which of
the local product managers she prefers to support in the development when trade-offs
need to be made [5]. On the other side, to increase the chances of being heard, each
local product manager should seek alternatives to product development with the
senior product manager, should act without considering other local product managers,
and should follow a constituent tactic by letting peripheral players with an indirect
stake lobby for the interests the local product manager is representing [5].
On the strategic level, design options should be identified and evaluated. The
considered designs should be meaningful for addressing the known communication
problems. The design that is selected should provide advantages compared with other
alternatives and have acceptable cost and impact for the concerned stakeholders.
Table 2 lists important options that can be derived from section 3.3.
Table 2: Evaluation of requirements communication network design options (SPM =
senior product manager, LPM = local product manager, PD = product development).
Design Change Advantages Disadvantages, Risks
Node + Link Properties No structural changes needed.
Strengthen status of SPM (intrinsic May increase capacity and May be just symptom
power). efficiency of the SPM. control.
Increase staff available to LPM and Addresses fundamental capacity Costly in long-term.
SPM. problems.
SPM regularly travels to LPMs for Increased efficiency. Useful for Reduced availability for
requirements elicitation purposes. building trust. PD. Costly.
SPM negotiates novel product concepts Useful for building trust. Supports Product concepts may
from PD with LPMs. technology innovation. May be irrelevant for LPMs.
reduce network latency.
31 Proceedings of the first International Workshop on Software Ecosystems 2009
Network structure
Direct contact of SPM to markets. Link between LPM and SPM Increases SPM
Markets
Market A
Competing Suppliers removed. May reduce network workload. SPM needs
Customer
Company 2 latency. SPM power increased. to travel more.
Company 1
Customer
Senior
Market B Product
Manager
Customer Development
Customer Team
Direct contact of LPM to development. Link between LPM and SPM Increases PD workload.
Markets
Market A
Competing Suppliers removed. May reduce network PD needs to travel
Customer
Company 2
Company 1 latency. PD power increased. more.
Customer
Local Product
Manager Development
Market B
Team
Customer Local Product
Manager
Customer
Split development team. Link between LPM and SPM Synergies lost (may be
removed. May reduce network addressed by a product
latency. May increase network platform team).
reliability.
Focus development on single market. Link between LPM and SPM Business volume
removed. Network latency and decreased.
load reduced. Power of Market A
increased.
Report to Steering Committee Conflicts between LPM and SPM May decrease SPM
can be escalated. and LPM capacity. May
increase network
latency.
Validation aims at verifying advantages, limitations and risks of selected RCN
changes and assuring that the changes are acceptable to stakeholders. For this
purpose, the changes are piloted in circumstances that are representative for the
concerned ecosystem. The validation results are analyzed by the process stakeholders.
If the validation results are acceptable, the process change is rolled-out on large scale.
If not, experimentation continues. The validation results, further, are used for
improving future predictions of impact of the various network design decisions
Roll-out aims at institutionalizing the validated changes to the RCN. Upon
successful roll-out, the real network corresponds to the planned one. During roll-out,
new challenges may be identified and used to launch a new improvement
development cycle.
32 Proceedings of the first International Workshop on Software Ecosystems 2009
5 Summary and Conclusions
Generally many stakeholder need to collaborate for bringing new products and
systems to success. These stakeholders pursue interests that they negotiate and agree
with interdependent stakeholders in an attempt to influence development efforts.
Important agreements are documented in the form of specifications.
This paper introduced requirements communication networks for describing and
analyzing such an ecosystem. Concepts from the body of knowledge of computer
networks were used to characterize node and link characteristics, network structure,
and end-to-end requirements communication network properties. The paper showed
how network analysis and specification supports process development for improving
requirements communication on a strategic level. An industrial exemplar has been
used to explain how the language is employed for designing and evaluating end-to-
end requirements communication networks with stakeholders that do not necessarily
stand in direct contact with each other. Network design options have been discussed
that allow evaluating changes to an existing network and capturing experience from
piloting and using a changed requirements communication strategy.
The presented approach enables better understanding of collaboration in a software
ecosystem by focusing on the relationships between interdependent software
stakeholders that need to agree with each other for building accepted products. It
shows how the bodies of knowledge of integrative negotiations and of network theory
can support analysis and design of stakeholder networks of software products. The
approach is used to describe snapshots of an evolving stakeholder network and to plan
collaboration among stakeholders by defining how they align their interests with
agreements. Not addressed has been tool support such as the use of modeling,
groupware, and communication technologies.
Additional research is needed to understand how networks can be modeled when
stakeholders belong to more than one group, i.e. when stakeholders are not
hierarchically organized. Currently, one diagram needs to be created per group
membership of one stakeholder. Pragmatic use of the diagrams eases this problem.
Additional research is also needed to better understand the effect of node and link
characteristics and of network structure on network properties. Empirical research can
improve current understanding of capacity, efficiency, load, latency, and reliability.
Such knowledge is necessary for building predictive models that assist evaluation of
network design options. Additional research, finally, is needed to evaluate the impact
of a process development approach based on requirements communication network
modeling on a software ecosystem.
References
1. Paech, B., Dörr, J., Koehler, M.: Improving Requirements Engineering
Communication in Multiproject Environments. IEEE Software 22 (2005) pp.40-47.
2. Regnell, B., Brinkkemper, S.: Market-Driven Requirements Engineering for
Software Products. In: Aurum, A., Wohlin, C. (eds.): Engineering and Managing
Software Requirements. Springer (2005) pp.287-308.
33 Proceedings of the first International Workshop on Software Ecosystems 2009
3. Stapel, K., Knauss, E., Allmann, C.: Lightweight Process Documentation: Just
Enough Structure in Automotive Pre-Development. 15th European Conference on
Software Process Improvement (EuroSPI 2008) (2008).
4. Grünbacher, P., Seyff, N.: Requirements Negotiation. In: Aurum, A., Wohlin, C.
(eds.): Engineering and Managing Software Requirements. Springer (2005) pp.143-162.
5. Fricker, S., Grünbacher, P.: Negotiation Constellations - Method Selection
Framework for Requirements Negotiation. Intl. Working Conference on
Requirements Engineering: Foundation for Software Quality (RefsQ'08) (2008).
6. Jansen, S., Brinkkemper, S., Finkelstein, A.: Providing Transparency in the
Business of Software: A Modeling Technique for Software Supply Networks. Virtual
Enterprises and Collaborative Networks (2007).
7. Yu, E.: Towards Modelling and Reasoning Support for Early-Phase Requirements
Engineering. 3rd IEEE Intl. Symposium on Requirements Engineering (RE'97)
(1997).
8. Gordijn, J., Yu, E., van der Raadt, B.: e-Service Design Using i* and e3value
Modeling. IEEE Software 23 (2006) pp.26-33.
9. Alexander, I., Robertson, S.: Understanding Project Sociology by Modeling
Stakeholders. IEEE Software 21 (2004) pp.23-27.
10. Kotonya, G., Sommerville, I.: Requirements Engineering with Viewpoints.
Software Engineering Journal 11 (1996) pp.5-18.
11. Bannerman, P.: Software Development Governance: A Meta-Management
Perspective. ICSE Workshop on Software Development Governance (2009).
12. Fricker, S., Gorschek, T., Myllyperkiö, P.: Handshaking between Software
Projects and Stakeholders Using Implementation Proposals. Intl. Working Conference
on Requirements Engineering: Foundation for Software Quality (RefsQ'07) (2007).
13. Pohl, K., Böckle, G., van der Linden, F.: Software Product Line Engineering:
Foundations, Principles and Techniques. Springer (2005).
14. Lewicki, R., Barry, B., Saunders, D.: Essentials of Negotiation. McGraw Hill
Higher Education (2006).
15. Akyildiz, I., Wang, X.: Wireless Mesh Networks. Wiley (2009).
16. Tanenbaum, A.: Computer Networks. Prentice Hall (2002).
17. Simon, H.: The Sciences of the Artifical. The MIT Press (1996).
18. Regnell, B.: Can We Beat the Complexity of Very Large-Scale Requirements
Engineering? Requirements Engineering: Foundations for Software Quality
(RefsQ'08) (2008).
19. Damian, D., Zowghi, D.: RE Challenges in Multi-Site Software Development
Organisations. Requirements Engineering 8 (2003) pp.149-160.
20. Fricker, S., Gorschek, T., Glinz, M.: Goal-Oriented Requirements Communication
in New Product Development. 2nd International Workshop on Software Product
Management (IWSPM'08) (2008).
21. Shewhart, W., Deming, W.: Statistical Method from the Viewpoint of Quality
Control. Courier Dover Publications (1986).