Dynamic application composition for distributed systems

大多数人使用的软件被设计和打包为在单个设备上运行(例如，在个人计算机上运行的电子表格)。这种方法的例外通常限于不在传统台式机器或个人计算设备(例如，蜂窝电话、PDA等)上运行的高端科学和商业应用。这种“标准”编程方法隐含地假设了一个基本的计算模型，其中所有的工作都在一台通用计算机上完成。然而，随着新的移动和嵌入式设备的普及，这种模式正在悄然消退。越来越多地，这种相互连接的设备协同参与计算。在这种现象的众多例子中，有对等文件共享应用程序和多用户游戏(大规模)，以及在iPod、手机等上运行的多媒体工具(小规模)。在这项提议中，我认为继续假设程序应该是在一台机器上运行并授权给特定个人的单一实体，这是过于严格的，应该被取代。我建议一种基于可执行组件(即执行自包含功能的可运行程序片段)的替代模型，这些组件可以动态地“组合”以创建根据特定用户和环境特征定制的应用程序。这种动态构建的应用程序可以在范围广泛的互连设备上和跨各种互连设备更高效地运行。例如，考虑这样一个场景：用户在旅行时需要使用功能有限的移动设备编辑文档。正常情况下，用户必须支付并预装一个大型、复杂且昂贵的应用程序，以支持在移动设备上编辑必要的文档。在我提出的方法中，只有软件的必要部分将按需动态组成，以便在移动设备上运行(如果必要，还可以在附近一台或多台提供计算服务的机器上运行)。此外，合成的软件可以进行量身定做，以适应设备无线连接互联网的小屏幕尺寸和有限的可靠性。

The software that most people use is designed and packaged to run on a single device (e.g. a spreadsheet running on a personal computer). Exceptions to this approach are normally restricted to high-end scientific and business applications that do not run on conventional desktop machines or personal computing devices (e.g. cell phones, PDAs, etc.). This "standard" programming approach implicitly assumes an underlying model of computing where everything is done on a single, general purpose computer. This model, however, is quietly fading as new mobile and embedded devices become prevalent. Increasingly, such interconnected devices are collaboratively involved in computations. Among the multitude of examples of this phenomenon are Peer-To-Peer file sharing applications and multi-user gaming (at the large scale) and multimedia tools running on iPods, cell phones, etc. (at the small scale).   In this proposal, I argue that continuing to assume that programs should be monolithic entities running on a single machine and licensed to specific individuals is overly restrictive and should be replaced. I suggest an alternative model based on executable components (i.e. pieces of runnable programs performing self-contained functions) that may be dynamically "composed" to create applications on the fly that are tailored to specific user and environmental characteristics. Such dynamically constructed applications can be run more efficiently both on and across a wide range of interconnected devices. For example, consider a scenario where a user needs to edit a document while travelling using a mobile device with limited capabilities. Normally the user would have to pay for and pre-install a large, complex and costly application to support editing the necessary document on the mobile device. In my proposed approach, only the necessary parts of the software would be dynamically composed on demand to run on the mobile device (and, if necessary, on one or more nearby machines offering compute services). Further, thecomposed software could be tailored to suit the small screen size and limited reliability of the device's wireless connection to the Internet.