Abstract
OMTP (Open Mobile Terminal Platform) is a global forum established by telecommunications providers to promote user-oriented mobile services and data business. Devised by OMTP, BONDI is a browser-based application or a mobile web run-time platform designed to help widgets utilize mobile device functions securely. BONDI enables applications programmed with web standard technologies—such as HTML, JavaScript, CSS, and AJAX—to access the internal functions of mobile devices. Because BONDI provides a standard way to reach internal resources, it allows for cross-platform development regardless of the operating system. However, web-based widgets are often vulnerable to network environments and can suffer from slow execution speeds as they become more complex. This study suggests methods to effectively operate and manage OMTP BONDI web widget resources and provides improved results based on performance evaluation experiments. The research focuses on enhancing module-loading speeds and overall operating time by ensuring only indispensable modules are loaded during execution.
Introduction: The Challenge of Mobile Web Widgets
In the early 2010s, mobile web widgets were viewed as a promising solution for cross-platform application development. The OMTP BONDI standard was specifically created to bridge the gap between web technologies and native device features like the camera, GPS, and messaging. Despite this potential, web widgets faced a significant hurdle: they were notably sluggish compared to native applications.
The standard BONDI runtime environment often suffered from inefficiencies. By default, engines would load unnecessary modules, and because widgets were distributed as compressed packages (.wgt files), devices had to decompress them every time they were launched. Furthermore, limited device memory often forced users to delete widgets, leading to repeated downloads and increased data usage. This paper addresses these bottlenecks through architectural and procedural optimizations.
The BONDI Platform Architecture
To understand the proposed optimizations, one must first look at the BONDI stack. The platform operates as a layer between the web engine (like WebKit or Gecko) and the device's operating system. It manages security through Javascript API Access Control and Device Capability Access Control, mediating how web code interacts with sensitive hardware features.
Proposed Optimization Methods
The researchers introduced four novel approaches to streamline widget performance and management:
1. Selective Module Loading
Standard runtimes are often too "eager," initializing every available module just in case a widget might need it. The authors proposed adding a loaded attribute to the <feature> tag in the widget's config.xml file. This allows the engine to load only the specific device APIs required by that widget, significantly reducing initialization overhead.
2. Redefined Widget Resource List (WRList)
Instead of parsing a widget's metadata from inside its compressed package at every launch, the system now parses this information once during installation. This metadata is stored in a fast-access "Widget Resource List" (WRList), allowing the engine to skip redundant parsing steps during the startup sequence.
3. Virtual Widget Cache
To eliminate the "decompression tax," the system implements a virtual cache. Frequently used widgets are kept in an uncompressed state in a temporary storage area, such as an SD card or internal SDRAM. This allows the engine to load the index.html and associated assets directly without waiting for a decompression algorithm to finish.
4. Widget Box
To solve the storage vs. accessibility dilemma, the authors devised a "Widget Box." This is a remote storage solution on a provisioning server where users can "park" unused widgets. This saves local device space while preserving user settings and avoiding the need to re-search for the widget in an app store later.
Key Results and Performance Evaluation
The researchers evaluated these methods using 10 real-world sample widgets from the OMTP "Widget Gallery," including a GPS Monitor and an SMS sender, tested on a Motorola XT720 smartphone. The results showed dramatic improvements:
- Startup Speed: The Virtual Widget Cache was the most impactful optimization. For the "GPS Monitor" widget, execution time plummeted from 2,583ms down to just 230ms—a more than 10x speedup.
- Module Efficiency: For the "Ajax demo" widget, the time spent loading modules dropped by nearly 87% (from 43.6ms to 5.5ms) due to selective loading.
- Combined Impact: Even the heaviest widgets, such as a Bike Sharing app, saw their startup times cut by more than half, moving from a 5-second wait to approximately 2.5 seconds.
Conclusion
This study demonstrates that the perceived performance gap between web widgets and native apps can be significantly narrowed through intelligent resource management. By moving computational costs to the installation phase and implementing a strategic cache, the authors achieved near-native startup speeds for web-based tools. Future work will focus on refining cache replacement algorithms to manage storage more intelligently based on usage frequency and last-access time, ensuring a balance between high-speed execution and device resource constraints.
Summary updated