Abstract: An intermediary system operates as an intermediary between content servers and user devices, and provides services for improving page load times as seen by end users. One such service involves converting a retrieved content page (e.g., web page) into a number of encoded layers that can be decoded and rendered by the user device. The intermediary system determines a suitable encoding technique for each of the layers. The intermediary system sends the encoded layers to a browser component running on the user device for rendering by the GPU.

Abstract: Techniques and solutions for providing a cloud browse service are described. For example, a client can request a web page. In response to the request, the client can receive a processed layer tree representing the web page. The processed layer tree can be created by a server environment (e.g., by creating an original DOM from obtained HTML and associated web page resources for the web page and creating the processed layers from the original DOM). The client can create a simplified DOM from the received processed layers and display the web page using the simplified DOM. Techniques and solutions for providing a browser as a service are described. For example, a web browser component can receive a processed layer tree representing a web page, create a simplified DOM, and display the web page.

Abstract: An intermediary system operates as an intermediary between content servers and user devices, and provides services for improving page load times as seen by end users. One such service involves converting all or a portion of a retrieved content page (e.g., web page) into a stream of graphics commands, such as OpenGL commands, that can be executed by the user device's graphics processing unit (GPU). The intermediary system sends these commands to a browser component running on the user device for execution by the GPU.

Abstract: Techniques and solutions for predicting web pages are described. Web page prediction can be performed using prediction models, including aggregate prediction models and user-based prediction models. Prediction models can be used to predict which web page (or which web pages) a user is likely to visit (e.g., to visit next after a current web page). Predicted web pages can be obtained in advance (e.g., pre-fetched and/or pre-rendered). Web page prediction can be performed by server computing environments and/or by client computing devices.

Abstract: Features are disclosed for determining which content item or items are likely to be requested by a particular user or device, or by a group of users or devices. The determined content items may be obtained independently of a request from the users or devices, and substantially current versions of the content items may be maintained at a server system, such as an intermediary system. Visual representations of the content items may be generated to reduce processing requirements at a user device. When the content items are determined to be likely to be requested by a particular user, a predicted time may also be determined, and the visual representations may be generated such that they are available at the predicted time.

Abstract: A multi-process browser architecture is provided that splits the browser process and the renderer processes. The browser process may run on a user device and the renderer processes may run on a renderer server accessible by the user device. The browser process running on the user device can direct the renderer server to generate a new renderer process each time a user elects to open a new browser window or tab. The renderer process running on the renderer server can determine what content resources are needed to render a page and request the resources. Once the content resources are obtained, the appropriate renderer process can begin generating a layout of the page. The renderer process may transmit the layout to the browser process such that the page can be displayed on the user device.

Abstract: Features are disclosed for automatically generating content requests and recording metrics and other information about execution of the requests. The requests can be background requests, executed by user computing devices during idle periods and without any user initiation. The background requests may be browser requests for content pages. Metrics and other information about execution of the requests may be recorded during execution of the background requests and reported to a performance analysis system. Instructions to execute background requests can be generated so as to develop a data set, such as a data set for analyzing request execution performance.

Abstract: Techniques and solutions for providing a cloud browse service are described. For example, a client can request a web page. In response to the request, the client can receive a processed layer tree representing the web page. The processed layer tree can be created by a server environment (e.g., by creating an original DOM from obtained HTML and associated web page resources for the web page and creating the processed layers from the original DOM). The client can create a simplified DOM from the received processed layers and display the web page using the simplified DOM. Techniques and solutions for providing a browser as a service are described. For example, a web browser component can receive a processed layer tree representing a web page, create a simplified DOM, and display the web page.

Abstract: Server systems are disclosed that receive content request and respond with hardware-independent graphics commands instead of, or in addition to, unprocessed content items. Some server systems may act as intermediaries between user devices and content sources. The server systems can obtain content pages from content sources, process the content pages (e.g., using a headless browser), and provide hardware-independent graphics commands (e.g., vector graphics commands, that do not require specialized graphics hardware for execution) to user devices in response to requests for content. The hardware-independent graphics commands can be executed by client browsers to generate a representation of the content page that is the same or substantially the same as the representation that would have been generated by processing the content pages locally.

Abstract: Server systems are disclosed that receive content request and respond with hardware-independent graphics commands instead of, or in addition to, unprocessed content items. Some server systems may act as intermediaries between user devices and content sources. The server systems can obtain content pages from content sources, process the content pages (e.g., using a headless browser), and provide hardware-independent graphics commands (e.g., vector graphics commands, that do not require specialized graphics hardware for execution) to user devices in response to requests for content. The hardware-independent graphics commands can be executed by client browsers to generate a representation of the content page that is the same or substantially the same as the representation that would have been generated by processing the content pages locally.

Abstract: Server systems are disclosed that receive content requests and respond with hardware-independent graphics commands instead of, or in addition to, unprocessed content items. The server systems can also generate text information regarding text in the requested content items, and provide the text information to user devices so that the user devices can have knowledge of the text in the content item. The user device can use the text information to handle user interactions with the content item, including copy, paste and search commands and other similar commands. Accordingly, the hardware-independent graphics commands-based representation may provide text interactivity and effects not otherwise available to content representations based on graphics commands or images of content.

Abstract: Server systems are disclosed that receive content requests and respond with hardware-independent graphics commands instead of, or in addition to, unprocessed content items. The server systems can also generate text information regarding text in the requested content items, and provide the text information to user devices so that the user devices can have knowledge of the text in the content item. The user device can use the text information to handle user interactions with the content item, including copy, paste and search commands and other similar commands. Accordingly, the hardware-independent graphics commands-based representation may provide text interactivity and effects not otherwise available to content representations based on graphics commands or images of content.

Abstract: Features are disclosed for determining which content item or items are likely to be requested by a particular user or device, or by a group of users or devices. The determined content items may be obtained independently of a request from the users or devices, and substantially current versions of the content items may be maintained at a server system, such as an intermediary system. Visual representations of the content items may be generated to reduce processing requirements at a user device. When the content items are determined to be likely to be requested by a particular user, a predicted time may also be determined, and the visual representations may be generated such that they are available at the predicted time.

Abstract: Features are disclosed for determining which content item or items are likely to be requested by a particular user or device, or by a group of users or devices. The determined content items may be obtained independently of a request from the users or devices, and substantially current versions of the content items may be maintained at a server system, such as an intermediary system. Visual representations of the content items may be generated to reduce processing requirements at a user device. When the content items are determined to be likely to be requested by a particular user, a predicted time may also be determined, and the visual representations may be generated such that they are available at the predicted time.

Abstract: Techniques and solutions for predicting web pages are described. Web page prediction can be performed using prediction models, including aggregate prediction models and user-based prediction models. Prediction models can be used to predict which web page (or which web pages) a user is likely to visit (e.g., to visit next after a current web page). Predicted web pages can be obtained in advance (e.g., pre-fetched and/or pre-rendered). Web page prediction can be performed by server computing environments and/or by client computing devices.

Abstract: Features are disclosed for determining preferred content request modes for client computing devices when initiating content requests. The request modes may correspond to direct requests (e.g., requests made from a client device directly to a content sever hosting requested content) or to indirect requests (e.g., requests made from the client device to the content server via an intermediary system). The preferred request modes made be based on a statistical analysis of performance data (e.g., prior content load times) obtained from one or more client computing devices for a given content item, group of content items (e.g., domain), and the like.

Abstract: Features are disclosed for selecting preferred content request modes on a client computing device when initiating content requests. The request modes may correspond to direct requests (e.g., requests made from a client device directly to a content sever hosting requested content) or to indirect requests (e.g., requests made from the client device to the content server via an intermediary system). The preferred request modes made be based on a statistical analysis of performance data (e.g., prior content load times) observed or recorded by the client computing device in connection with prior content requests. Randomly selected request modes may be used to provide additional data for performance analysis.

Abstract: Features are disclosed for generating request decision models for use by client computing devices to determine request paths or modes for content requests. The request modes may correspond to direct requests (e.g., requests made from a client device directly to a content server hosting requested content) or to indirect requests (e.g., requests made from the client device to the content server via an intermediary system). The request decision models may be trained by a machine learning algorithm using performance data (e.g., prior content load times), contextual information (e.g., state information associated with devices at times content requests are executed), and the like.

Abstract: Processes are disclosed for identifying site-specific or domain-specific URL parameters that are likely associated with personally identifiable information (PII). The processes involve analyzing the browsing session records of multiple users to assess, for a given site or domain, the uniqueness levels of particular URL parameters. These uniqueness levels, alone or in combination with other criteria, are used to determine whether each such parameter is likely associated with PII. The results of this analysis may be used to filter likely PII from session records, such as by modifying and/or deleting the URLs likely to contain PII.