Abstract
As the Internet transitions from a network of data to a network of functionalities, the challenge of cross-application integration becomes increasingly complex. Current integration paradigms often suffer from a "responsibility imbalance," placing too much burden on either providers or end-users. This paper proposes a Multiple Participant (MP) cooperation model that distributes the effort of integration among application providers, end-users, and a central intermediary. By enabling a "Federated Internet of Applications," the MP model achieves higher task coverage for long-tail user needs while maintaining manageable preparation requirements for all stakeholders.
Introduction
Modern users increasingly require their applications to work in concert—for instance, automatically updating a delivery address in a retail app based on a hotel booking in a travel app. However, existing integration methods face significant hurdles:
- Provider Development (PD): Large providers cannot natively integrate with every niche application, leading to low task coverage for personalized needs.
- End-user Development (EUD): Tools like IFTTT empower users but impose a heavy technical burden, often failing when APIs are incompatible or not exposed.
The authors argue that no single participant can comprehensively manage integration. Instead, a balanced model is required to optimize effort allocation without overwhelming any single party.
The Multiple Participant (MP) Model
The proposed MP model introduces a hybrid architecture involving three primary roles:
- Applications: Autonomous entities that provide specific functionalities and interfaces.
- Users: The beneficiaries who define their personal "userspace"—a private network of application connections tailored to their needs.
- Intermediary: A centralized service consisting of an
AppMarketandAppNetRuntimethat coordinates interactions.
Operational Stages
The model operates in two distinct phases:
- Configuration Stage: Providers publish interfaces to the
AppMarket. The system then recommends potential connections to users based on their active applications. Users confirm these connections to build their unique userspace. - Execution Stage: When a trigger occurs, the
AppNetRuntimeutilizes an extended OAuth protocol to identify user preferences and facilitate communication. Data incompatibility is managed via a named message and adapter mechanism, where adapters transform data into the specific formats required by receiving applications.
Evaluation and Results
The model was evaluated through a qualitative case study involving a cross-application scenario: a user ("Bob") whose hotel reservation automatically updates his shipping address for a retail purchase. The authors compared the MP model against traditional EUD and PD approaches.
As shown in the comparison, the MP model achieves "Very High" task coverage by supporting long-tail requirements that providers typically ignore, while keeping the preparation effort "Moderate" for both developers and users. This balance addresses the scalability issues inherent in PD and the complexity barriers of EUD.
Key Insights
- The "Userspace" Concept: By shifting from a global web of apps to a personalized userspace, the model allows for user-specific logic without requiring applications to have prior knowledge of one another.
- Balanced Effort: Providers define what they can do (interfaces), users decide who interacts, and the intermediary manages how they coordinate.
- Decoupled Transformation: Moving data transformation to specific adapters rather than a central bus enhances system scalability.
Limitations and Future Work
While promising, the MP model requires a paradigm shift toward cooperative, interface-first application development. The reliance on a centralized intermediary also raises questions regarding security, privacy, and potential bottlenecks that require further exploration. Future research will focus on broader empirical testing with real-world users and a more diverse set of applications to validate the scalability of the adapter mechanism.
Summary updated