【正文】 del2 機制主要滿足了開發(fā)商對要求 Web應用邏輯和用戶界面獨立的需求。特別是 ,就 Servlet 當控制器 ,不僅控制選擇同時響應模型的模塊 ,還負責對處理 JSP 轉移數(shù)據(jù)提供模塊。因此 ,在下面的網(wǎng)絡應用系統(tǒng)框架 ,我們會進一步劃分出功能的控制器。這三個控制對象 /類根據(jù)原始的控制器劃分。 首先 ,我們考慮 SYSController和系統(tǒng)的描述數(shù)據(jù)庫。系統(tǒng)描 述數(shù)據(jù)庫可以是一個 XML[2]文件或一套XML文件。執(zhí)行邏輯主要包括互動模式、數(shù)據(jù)來源與異常處理等，這些組織在一個樹形結構中。它結合了用戶的要求和來自數(shù)據(jù)的執(zhí)行邏輯 ,然后把它們交給傳輸器來封裝。 其一是使用 DOM對象(Document Object Model)[3]。數(shù)據(jù)庫將保留在內(nèi)存中的 DOM樹結構。兩者有各自的優(yōu)缺點。它可以在大型數(shù)據(jù)庫上使用 ,但是會暫用更多的系統(tǒng)資源。唯一的缺點是它獲得信息比 DOM難。該層通過調(diào)用給定模塊完成定義執(zhí)行邏輯。這些對象和模塊根據(jù)特定的執(zhí)行邏輯封裝響應方法。然后按照給定的 數(shù)據(jù)格式封裝模塊 ,冰傳輸給 LPM中心。 最后 ,我們研究一下 LPM中心和 LPM數(shù)據(jù)庫。在另一方面 ,有一個網(wǎng)頁可有一個邏輯頁或多條邏輯頁。因此事實上 ,LPM是作為標簽庫 (JSP)或數(shù)據(jù)捆綁 (.)而開發(fā)的。在這部分 ,LPM中心從模型中心收到數(shù)據(jù)。 注意到這 3個控制對象 ,除了 SYSController,另外兩個不是直接進入自己的核心數(shù)據(jù)庫。它實現(xiàn)了通過 ModelFactory的存取。在在系統(tǒng)的開發(fā)時期各部分的開發(fā)者三者冰不具有相互的干涉。在過去這對網(wǎng)絡應用程序的開發(fā)來說是不可能的。所以它可以明顯提高整個系統(tǒng)的可擴充性。 檢查目前的網(wǎng)絡應用程序設計模式 (如 CGI,JSP Model2等等 )，我們可以獲得初步結果： MVC模式是一個優(yōu)于其他系統(tǒng)邏輯框架的網(wǎng)站應用系統(tǒng)開發(fā)模式，它平衡了合作開發(fā)、系統(tǒng)的可擴展性和可維護性。例如，在通過 CGI開發(fā)的網(wǎng)頁系統(tǒng)中， CGI程序不僅需要處理執(zhí)行邏輯 ,同時也負責系統(tǒng)報告的格式。此外，它甚至會使系統(tǒng)的邏輯層次混亂。使用服務器腳本語言 (如 ASP,PHP等等 )實現(xiàn)的網(wǎng)頁應用系統(tǒng)也不符合要求。由此JSP Model2的不足已在上文描述過了 ,所以沒有這個必要在這贅述了。改進后的 MVC設計模式框架比傳統(tǒng)方法有更清楚的結構，每個模塊間的聯(lián)系甚少。它將執(zhí)行邏輯從表現(xiàn)模式獨立出來，實施并行進行發(fā)展。本篇討論的框架已經(jīng)在 UEST網(wǎng)絡應用系統(tǒng)上得以實施 參考文獻 [1] Erich Gamma. Design Patterns. . AddisonWesley Pub Co. [2] Didier Martin. Professional XML. . Wrox Press. [3] Stephen T. Mohr. Designing Distributed Applications with XML, ASP, IE5, LDAP amp。 Wang Yanhua, the Master of the UEST of China。 Han Hong, the . of UEST Of China model to provide different presentations. You can also create new views for a model without rewriting it. 2. Background and Problems Even though the MVC design pattern was already put forward in smalltalk80 and widely used in software design, it is difficult to introduce it into the Web application development. It was mainly due to the following two factors: 1) First, it was difficult to separate the programming language from HTML. Early CGI programs created the HTML content by using character string output. In nature, it divided the HTML content into several portions, and then inserted them into the CGI programs. After the appearance of Script languages, the way changed to insert the programs written by script languages into HTML content. However, these two ways had a mon deficiency that they could not separate the programs from the HTML content absolutely. Therefore, it is very difficult to implement the MVC detached threelayer pattern based on these two ways. 2) Second, functions of the script languages are less weak. They lack some necessary technologies to support the MVC design pattern, such as message mechanism and event response mechanism. This situation did not get any improvement until the release of the JSP Model 2[4]. Depending on the powerful functions of Java language, the JSP Model 2 implements the MVC design idea to some extent. Figure 1 shows the mechanism: ? Model (the data layer): it is encapsulated into the Enterprise JavaBean. ? View (the presentation layer): JSP is responsible for the presentation of pages, such as the presentation format, paging, etc. ? Controller (the control layer): Servlet receives the user input in pages and executes the relevant transaction (calling the corresponding EJB modules), then gives the results to the relevant View, so JSP can present these results. This mechanism of JSP Model 2 basically meets the developers’ requirements to detach the Web application transaction logic from the user interface. But some limits still exist. For example, although JSP Model 2 has implemented the basic idea of the MVC design pattern, the relationship of these three layers (Servlet, JavaBean/EJB and JSP) is still close. Especially the Servlet, which acts as the Controller, not only controls the selection and call of Model’s modules, but also is responsible for transferring data provided by modules to the corresponding JSP to handle. Because of this process, Servlet has to implement very plicated transaction logic by itself, which is a tough problem for the system development and maintenance. Thus, in the following Web application system frame, we will further divide the functions of the Controller. 3. Solution As show in Figure 2, this system frame contains three kernel control objects/classes (SYSController, Model Center, and LPM (Logical Page Module) Center) and their own kernel database (system description database, model database, and LPM database). These three control objects/classes are divided from the raw Controller. Using these control objects/classes and databases, the following parts of this paper will explain the design idea and implementation mechanism of this system frame. Firstly, we consider the SYSController and the System Description Database. They are responsible for receiving the user requests and searching the relevant transaction logic according to these requests. The System Description Database can be a XML[2] file or a set of XML files. This database describes the different transaction logic related to different user requests one by one. The transaction logics mainly include the interactive mode, data sources, and error handling, etc. They are anized in a tree structure. The SYSController searches the System Description Database. It bines the user requests and the transaction logic searched out from the database, and then delivers them to Messenger to encapsulate. There are two ways for SYSController to search the database. One uses the DOM (Document Object Model) object[3]. When the application system boots, this way requires the whole database to be loaded into RAM after analyzed by the DOM Parser. The database resides in the RAM in a DOM tree structure. The other one uses SAX (Simple API for XML) interface. Both of them have their own advantages and disadvantages. Using DOM can implement structural search. It can achieve the efficiency in the case of larger database, but it al