TL;DR
This paper tackles the challenge of providing a good web browsing experience on resource-constrained smartwatches. They propose a "fast access browsing" model that leverages lightweight "widget views" for quick information retrieval, similar to notifications, and a dedicated wearable browser for richer content. This approach significantly reduces loading times and memory usage compared to traditional web browsing on wearables.
What Problem Does This Paper Address?
Smartwatches and wearable devices have become increasingly popular, offering convenient access to information and applications. However, a significant limitation is their ability to consume web content. This is primarily due to their inherent constraints: tiny displays, limited processing power, and battery life concerns. While notifications offer quick access to relevant information, running full-fledged web applications or traditional browsers on these devices is often a poor experience. Existing smartwatch platforms offer lightweight application models like watchOS Glances, Android Wear Always-on apps, and Tizen widgets, which provide fast access to information but are not designed for rich web browsing. This paper aims to bridge this gap by proposing a new model for web browsing on wearables that balances quick access with a more comprehensive browsing experience.
Methods
The authors propose a "fast access browsing" model with three core components:
- Mobile Browser: Runs on a connected smartphone, responsible for parsing web app manifests and initiating widget view installation on the wearable.
- Widget View Runtime: Manages the lifecycle of widget views on the smartwatch. This includes installation, execution, and updating of these lightweight web snippets.
- Wearable Browser Extension: A dedicated browser on the smartwatch designed for richer web content. It supports features like CSS snap scrolling for better readability on small screens and CSS media queries adapted for circular displays.
The core innovation lies in the widget view, which is a constrained web application designed for quick, always-on access. These widget views are built using a restricted set of HTML, CSS, and JavaScript, focusing on essential information display. The system also defines constrained web specifications for these widget views, limiting resource loading (e.g., no heavy external CSS, JS, or large images) and file sizes to ensure rapid loading and low memory footprint. A widget view validator is also developed to help developers adhere to these specifications.
Datasets
The paper analyzes approximately 90 smartwatch applications, specifically focusing on 45 Tizen widgets, 35 watchOS Glances, and 10 Android Wear Always-on applications. These applications are categorized into domains like health, planning, information, and control. The authors do not introduce new datasets but rather use existing smartwatch applications to derive requirements and evaluate their proposed system.
Key Insights and Ideas
- Widget Views as a Notification-like Experience: The concept of widget views, inspired by the rapid access of notifications, is a key insight. This allows users to get essential information without launching a full application.
- Constrained Web Specifications are Crucial: The paper highlights that simply using standard web technologies on wearables is not feasible. Defining strict constraints on HTML, CSS, and JavaScript is essential for achieving acceptable performance and resource usage.
- Separation of Concerns for Browsing: The model effectively separates the quick-access widget view from the richer, full-fledged wearable browser. This allows users to choose their preferred browsing experience based on their needs.
- Importance of Sandboxing: The implementation emphasizes the use of sandboxing for widget view clients to ensure system stability and security, preventing crashes in one widget from affecting others.
- Mobile Websites are Too Heavy for Wearables: The evaluation clearly demonstrates that conventional mobile websites are too resource-intensive for direct rendering on smartwatches, even with multi-process architectures.
Key Results
- Loading constrained web contents via widget views takes less than 200ms.
- Widget views consume only 26% of the memory footprint compared to a conventional web browser.
- The system can support up to 20 widget views on a smartwatch with 512MB of RAM.
- Loading sample web contents with constrained specifications takes less than 200ms on average, compared to 1200-1500ms for single- and multi-process WebKit.
- The accumulated memory usage for widget views is significantly lower, with widget view clients requiring no more than 5MB on average, compared to 6MB for single-process WebKit and 12MB for multi-process WebKit.
Limitations and Future Work
- Limited Web API Support: The constrained web specifications intentionally restrict certain web APIs (e.g., audio, video, heavy JavaScript) to meet performance goals. This limits the expressiveness of widget views.
- No Support for Rich Interactions in Widget Views: Widget views are primarily for static or simple data display, with interactions limited to tap events.
- Future Work: The authors plan to develop a JavaScript framework and a server-based pub/sub broker system to improve the reliability and performance of widget views with continuously updated content. They also aim to incorporate the concept of single-page applications into the smartwatch runtime environments.
Summary updated