【作者】 林彬；

【导师】 冯凤娟；

【作者基本信息】 北京交通大学 ， 软件工程（专业学位）， 2014， 硕士

【摘要】 随着科技的发展,3G通信技术和wifi越来越普及,而且手机、平板电脑等无线设备功能越来越强大,使用无线设备来浏览网页的人越来越多。这些无线设备分为以下几种：使用老式java系统的、使用智能操作系统但系统版本较低的、使用较高版本智能系统的、平板电脑等等。不同类型的设备,显示能力是不一样的,那么同一个网页如果想在所有的无线设备上都正常显示,只能使用最简单、最广泛兼容的内容,这样对于高级手机而言无疑浪费了手机软硬件能力,对于用户而言,展示的内容也不够丰富,尤其是平板电脑用户,视觉体验会更差。针对上述问题,本人实习所在公司实现了一套解决方法：即服务中间件首先对来自无线用户的网页请求进行解析,解析后可以得到用户无线设备的类型、展示能力的评级和用户无线设备的基本属性,如操作系统、版本号、浏览器类型、厂商、机型、接入方式等,将解析后的结果传给下游模块,下游模块根据这些结果选择网页模板,从而生成最适合在该设备上展示的网页,再将网页返回给用户。本文的内容是服务中间件的设计与实现,即如何解析来自用户的请求,对于每个请求如何进行分级,在请求中如何找到无线设备的属性等等。本项目分为三层：域名适配层,负责初步分析设备的类型,是手机、个人电脑还是平板电脑等；数据层,负责解析用户的设备属性,如操作系统及其版本号、浏览器及其版本号等；策略层：根据设备属性和设备类型给出对用户设备展示能力的评级,以供选择网页模板使用。本人的主要工作包括前期的数据准备和分析,服务的具体设计与实现,以及后期的效果评估和测试等,其中需求分析和数据分析是本人和产品经理共同完成的,项目的具体设计与实现是本人独立完成的,项目的测试是本人和测试人员一起完成的。在实现服务的过程中主要用到了如下技术：数字摘要算法、多模匹配算法、负载均衡算法等,为了提高项目运行的效率,本项目是使用C语言实现的,以函数调用的方式来完成项目的组建。目前本项目已经上线了,百度公司几乎所有的产品都调用了该服务中间件来对无线设备展示能力进行评级。在上线初期,项目出现了一些错误,经改正后达到了比较好的效果,目前对用户无线设备属性的识别率在97%以上,展示能力的评级的准确率在99%以上。项目上线后用户反响良好,用户数量和日均访问量出现了大幅增长。更多还原

【Abstract】 With the development of science and technology,3G communication technology and wifi are more and more popular, and the power of mobile phones or tablets are greatly improved. So there are more people suffer the internet by mobile phones. These wireless devices are divided into the following categories:using java system, using low-version smart phone operating system, using high-version smart phone operating system, tablet and so on. Different mobile phones have different display-power, so one same web page aimed to be normally displayed on all wireless devices must only use the content that is most simple and most widely compatible. For advanced mobile phone, this is a waste of the hardware and software capabilities, for the users, the web page is not rich enough, especially for the tablet-users, visual experience will be even worse.In response to these problems, the company where I practiced in implements a solution:set a middleware service to resolve the page-request from users at first, then we can get the type of the wireless device and we can also get some basic property such as operating system, version, vendor, model, browser, net-type and so on, so we can rate the display-power of the device. Then we pass all the resolved results to downstream module. The downstream module selects web page template depending on these results, so it can generate the most suitable web page for the device and then return the web page to the user. The content of this thesis is the design and the realization of the middleware service, including how to resolve the request of the user, how to divide the display-power of each request into several different levels, how to find the attribute of the wireless device in each request. The project is divided into three layers:domain adaption layer is in charge of resolving the type of the device, such as phone, personal computer and so on, data layer is in charge of resolving the basic attribute of the device, such as device os, browser type, strategy layer is in charge of deciding the level of the device’s display-power basing on the device’s attribute.My work includes the following parts of task:the preparation and analysis of data in prophase, the detail design and realization of the service and the effect evaluation in late stage. Among these tasks, request analysis and data analysis was completed by the product manager and me, the design and implementation was completed by myself, the test of the project was completed by quality assurance and me. During the realization of the service, such technologies are used, digital signature algorithm, multi-model match algorithm, load balancing algorithm and so on. To improve efficiency, the project is implemented in C language, the construction of the project is based on the call of functions.Now, the project is already online, almost all the product in Baidu company call the service to rate the display-power of the device. When the project was just launched, there were some mistakes, after I fixed them, the project has been running well until now. Now the rate of recognizing-attribute is above97%, the correct rate of the distinguishing-display-level is above99%, the project received a good response, the daily visits has been growing rapidly since launched.更多还原