Games and Information

【转载】数据挖掘技术研究进展作者：speedfish 日期：2004年06月14日

杨宁 — Wed, 14 Jul 2004 23:39:00 GMT

数据挖掘技术研究进展作者：speedfish 日期：2004年06月14日
--------------------------------------------------------------------------------　　

    1．前言
      随着计算机技术和信息技术的发展，信息的增长速度呈现指数上升，最近几十年产生了很多超大型数据库，遍及超级市场销售、银行存款、天文学、行政办公、科学研究、信息量的急剧增长，使传统分析方法远远不能满足现实的需求。面对海量数据，如何从中发现有价值的信息或知识，成为一项非常艰巨的任务。人们急切的需要一种去粗存精、去伪存真的技术，能够从海量的数据中提取知识和信息的数据挖掘技术应运而生[1]。
    数据挖掘(Data Mining)就是从大量的、不完全的、有噪声的、模糊的、随机的数据中，提取隐含在其中的、人们事先不知道的、但又是潜在有用的信息和知识的过程[2]。这些数据可以是结构化的,如关系数据库中的数据，也可以是半结构化的，如文本，图形，图像数据，甚至是分布在网络上的异构型数据。发现知识的方法可以是数学的，也可以是非数学的，可以是演绎的，也可以是归纳的。发现了的知识可以被用于信息管理、查询优化、决策支持、过程控制等，还可以进行数据自身的维护。数据挖掘借助了多年来数理统计技术和人工智能以及知识工程等领域的研究成果构建自己的理论体系，是一个交叉学科领域，可以集成数据数据库、人工智能、数理统计、可视化、并行计算等技术。
    2. 数据挖掘技术的发展
    　　1989年8月，在美国底特律召开的第11届国际人工智能联合会议的专题讨论会上首次出现KDD这个术语，随后在1991，1993，1994年都举行了KDD专题讨论会，集中讨论数据统计、海量数据分析算法、知识表示、知识运用等问题。KDD国际学术大会研究重点逐渐从发现方法转向系统应用，并且注重多种发现策略和技术的集成，以及多种学科之间的相互渗透，数据挖掘和知识发现成为当前计算机科学界的一大热点。1998年在美国纽约举行的第四届知识发现与数据挖掘国际学术会议上有30多家软件公司展示了数据挖掘软件产品，不少软件已经在北美和欧洲的国家得到应用。
    META Group曾做出这样的评论：“全球重要的企业、组织会发现，到21世纪数据挖掘技术将是他们商业成功与否的至关重要的影响因素”。IBM公司发布了基于标准的数据挖掘技术-IBM DB2智能挖掘器积分服务，可用于个性化的解决方案。两大统计软件公司SAS和SPSS也推出了各自的数据挖掘工具Enterprise Miner和Clementine[3]。此外，在Internet上还有不少KDD电子出版物，其中以半月刊Knowledge Discovery Nuggets最为权威，另一份在线周刊为DS(决策支持)，1997年开始出版。自由论坛DM Email Club可以通过电子邮件讨论数据挖掘和知识发现的热点问题。数据挖掘是数据库和信息决策领域的最前沿的研究方向之一，已引起了国内外学术界的广泛关注。在我国已经开始进行数据挖掘技术的研究，但还没有看到数据挖掘技术在我国成功应用的大型案例。

    3.数据挖掘的目的
    数据挖掘并不专用于特定领域，它需要凝结各种技术和创造力去探索可能隐藏在数据中的知识。在很多情况下，应用数据挖掘技术是为了实现以下三种目的：
    ·发现知识：知识发现的目标是从数据库存储的数据中发现隐藏的关系、模式和关联。例如,在商业应用中数据挖掘可用于发现分割、分类、关联、喜好四种知识。发现分割知识可以将客户记录分组，策划为客户度身定做的推销活动。发现分类知识可以将输入的数据分配到预定义的类别中，发现和理解趋势以及对文本文档的进行分类等。发现交叉销售的机会是一种关联知识，以及发现大部分客户的喜好的知识[4]。
    ·使数据可视化：分析人员需搞清楚数据库中存储的大量信息的含意。在做任何分析之前，需先将待处理的数据人性化，并寻找显示数据的好方法。
    ·纠正数据：在结合大规模的数据库时，数据库的数据常常是不完整的，而且通常包含错误和自相矛盾的信息。数据挖掘需要以最稳定的方法识别和纠正这些问题[5]。

4. 数据挖掘的方法

    数据挖掘方法分为统计方法、机器学习方法、神经网络方法和数据库方法。
    ·统计学的方法是数据挖掘的经典方法。统计方法中包括回归分析(多元回归、自回归等)、判别分析(贝叶斯判别、费歇尔判别、非参数判别等)、聚类分析(系统聚类、动态聚类等)、探索性分析( 主元分析法、相关分析法等)等[6]。
    ·机器学习中包括归纳学习方法(决策树、规则归纳等)、基于范例学习、遗传算法、粗糙集等。粗糙集能够对不确定、不完整信息的进行处理，而遗传算法具有全局最优搜索的能力[7]。
    ·神经网络方法具有处理非线性数据和含噪声数据的能力。神经网络的常用算法包括前向神经网络(BP算法等)、自组织神经网络(自组织特征映射、竞争学习等)等[8]。
    ·数据库方法主要是多维数据分析或OLAP方法[9]。OLAP系统的数据库为高效存储静态数据构建。其存储结构的设计是为了高效检索数据，尤其是聚合数据，比如求总和或是其他运算[10]。

5. 数据挖掘的主要步骤

    数据挖掘是一个复杂的过程，它的一般步骤是：
    （1）分析问题: 源数据数据库必须经过评估确认其是否符合数据挖掘标准。以决定预期结果，也就选择了这项工作的最优算法。
    （2）提取、清洗和校验数据：提取的数据放在一个结构上与数据模型兼容的数据库中。以统一的格式清洗那些不一致、不兼容的数据。一旦提取和清理数据后，浏览所创建的模型，以确保所有的数据都已经存在并且完整。
    （3）创建和调试模型：将算法应用于模型后产生一个结构。浏览所产生的结构中数据，确认它对于源数据中"事实"的准确代表性，这是很重要的一点。虽然可能无法对每一个细节做到这一点，但是通过查看生成的模型，就可能发现重要的特征。
    （4）查询数据挖掘模型的数据：一旦建立模型，该数据就可用于决策支持了。在微软的数据挖掘解决方案中，该过程通常使用VB或ASP通过OLE DB for Data Mining Provider编写前端查询程序[11]。
    （5）维护数据挖掘模型：数据模型建立好后，初始数据的特征，如有效性，可能发生改变。一些信息的改变会对精度产生很大的影响，因为它的变化影响作为基础的原始模型的性质。因而，维护数据挖掘模型是非常重要的环节[5]。

6. 数据挖掘的常用工具

    数据挖掘工具的市场一般分为三个组成部分：通用型工具、综合／DSS／OLAP数据挖掘工具和快速发展的面向特定应用的工具。
    通用型工具占有最大和最成熟的那部分市场。从定义上说，它们是非面向特定应用的适合于各种需要的情况，其中包括的主要工具有SAS Enterprise Miner、IBM Intelligent Miner、Oracle Darwin、SPSS Clementine、Unica PRW、SGI MineSet等软件。
    综合数据挖掘工具这一部分市场反映了商业对具有多功能的决策支持工具的真实和迫切的需求。商业要求该工具能提供管理报告、在线分析处理和普通结构中的数据挖掘能力。这些综合工具包括Cognos Scenario和Business Objects等。
    面向特定应用工具这一部分工具正在快速发展，在这一领域的厂商设法通过提供商业方案而不是寻求方案的一种技术来区分自己和别的领域的厂商。这些工具是纵向的、贯穿这一领域的方方面面，其常用工具有重点应用在零售业的KD1、主要应用在保险业的Option＆Choices和针对欺诈行为探查开发的HNC软件[12]。

7. 数据挖掘的应用领域

数据挖掘技术从一开始就是面向应用的，它不仅可用于特定数据库的简单检索查询调用，而且要对这些数据进行统计、分析、综合和推理，以指导实际问题的求解，发现事件间相互关联关系和进行预测。商业上通过使用面向CRM数据挖掘技术使公司可以快速的对顾客的需求做出反应，并且可以检验商业策略的成效[13]。在科学研究方面，一个天文学上的著名应用系统SKICAT就是相当成功的数据挖掘应用，利用该系统，天文学家已发现16个新的极其遥远的星群。在生物医学和DNA数据分析上，数据挖掘可以完成异构、分布式基因数据库的语义集成，用关联规则分析同时出现的基因序列，用路径分析发现在疾病不同阶段的致病基因等[14]。NBA教练就运用Advanced Scout来挖掘信息，安排阵型，提高了获胜的机率；在金融投资方面，FALCON系统是信用卡欺诈估测系统，已被相当数量的银行采用，FAIS是一个用于识别与洗钱有关的金融交易系统，LBS Capital Management则使用了专家系统、神经网络和基因算法技术来辅助管理多达6亿美元的有价证券；在电信业，利用数据挖掘可以确定潜在的盗用者和他们的非典型使用模式，检测想侵入用户账户的企图以及发现需要引起注意的异常模式，从而保证电信业的正常运转。在电力系统中，应用于数据挖掘技术的数据有电力系统的范围特征（包括是时间和空间上的）与统计特征，通常有几千个状态变量；混合存在的离散信息（如网络拓扑结构的改变或保护动作等）和连续信息（如某些连续变化的状态变量）；对某些不确定量的掌握和处理（如噪声和不完整信息等）[15]。在Internet，利用数据挖掘技术可以Web后台数据库进行分类处理，从而向用户提供更快、更准、更有效的信息。同时，也可通过对Web日志进行挖掘，发现Web用户访问模式，在高度相关的站点提供快速有效的访问通道，帮助更好地设计Web主页，改善市场营销决策[16]。

8．结束语
越来越多人们已经认识到数据挖掘技术能将原始数据转换为有意义的形式，从中挖掘潜在的巨大商业价值和所蕴含的科学知识，数据挖掘已成为电子商务淘金的首要的手段。具有影响的"MIT技术回顾"最近对数据挖掘技术极力赞扬，认为其已成为十大新兴技术之一[4]，将彻底改变数据库的处理模式。随着数据挖掘技术不断被应用到新的领域和各种算法不断被应用到数据挖掘领域中，将更大激发数据挖掘技术的潜力，进一步推进数挖掘技术的发展和普及[17]。

【转载】五大知识管理(KM)的产品与服务作者:孙定

杨宁 — Wed, 14 Jul 2004 23:10:00 GMT

五大知识管理(KM)的产品与服务

作者:孙定

知识管理解决方案的核心内容是制定知识管理策略。知识管理策略要解决观念问题，要突破信息时代形成的思维定式，更新知识，使观念向知识时代校正。知识管理策略还要解决机构的文化问题，使机构具有知识时代所要求的组织学习能力并建立知识共享机制。接下来是选择适当的产品，开发知识管理项目。因此，知识管理市场具有咨询服务需求与技术产品需求共生的特点。前者解决知识更新、观念更新、策略制定、文化改造、调整机制等问题；后者解决具体实现的问题。正是基于知识管理的这种特点，重要的知识管理供应商都同时提供咨询服务和技术产品。

　　知识管理产品与服务的另一特点是种类繁多，每个供应商都有自已的一套说法，这些说法互不相同，甚至差别巨大。这是由两个原因造成的：首先，目前无论在学术上还是在实际应用中，知识管理都处于非常早期的阶段，其定义有数百种，学术上也有很多不同的观点，供应商当然是各取所需；其次，供应商都是从自已原先的领域进入知识管理领域，拥有不同的技术和产品，而知识管理本身与其说是一种新技术不如说是一种新观念，大量现有产品与技术都与知识管理相关，供应商所做的只是根据知识管理的需求，重新定位现有的产品。

　　这里着重讨论一些重要的、具有不同特点的知识管理产品与服务。

　　Lotus:以专取胜

　　虽说Lotus与IBM本是一家，知识管理论调也一样，但各自有各自的知识管理产品，所以还是要分开说。

　　在所有知识管理解决方案厂商中，Lotus给人印象最为深刻。知识管理所必需的文档管理和群件技术在1998年前后已经是Lotus的主打产品。而Lotus Notes本身是一个可完成多种应用的平台，虽然不是浏览器界面，但在原理上已经很接近企业门户，这些都是Lotus进入知识管理市场的先天优势。这两年知识管理的兴起，对Lotus来说实在是一个天赐良机。Lotus在知识管理上狠下一番功夫，拼命赌一把也就在情理之中了。　　Lotus、IBM研究中心、IBM知识管理研究所共同对Lotus专业服务以及IBM全球知识管理服务机构在全球的2万个客户的知识管理实践进行了调查，以Lotus现有技术为基本出发点，制定出独特的理论框架，并确立了知识管理产品策略。第一个产品K-Station企业门户和其配套产品Discovery Server已经完成。

　　Lotus认为，仅仅将知识管理局限在从海量信息中提取有用资料是不够的，还要找到具有专业知识的人，这些人还要交流、互动、进行创造性的工作。于是，Lotus将数据、资料及处理过程定义为"事物（Thing）"、将建立在网上的虚拟工作环境定义成"场所（Place）"、将员工、客户、专家、合作伙伴等定义成"人（People）"，而在人、场所、事务之间建立有机关联才是理想的知识管理环境。

　其中，K-Station已经具有知识管理系统必备的知识管理功能，Discovery服务器则是对前者的增强。

　　在K-Station中，每个人都有自已的场所--个人场所（Personal Place）。个人场所为担任不同角色的人员提供定制的日常工作环境。在个人场所中可进行电子邮件处理、管理日程、讨论、获取订阅资料、编辑文档等操作。沟通场所（Community Place）为由相关人员组成的小组提供了共享与共同工作的环境。所有个人文档都被加上了基于场所的标签，并按场所将文档进行分类归档。这种机制为文档的共享和检索提供了方便。在场所中可以看到何人正在线上，并列出共享场所的清单，在线上的人可以相互进行即时的消息沟通。目前，K-Station必须在Domino环境下运行，因此系统中至少要有一个Domino服务器。

　　微软：追求通俗

　　微软一方面将现有产品基本上都贴了知识管理的标签，一方面也在开发新一代知识管理产品。微软的新一代知识管理产品正在进行第三版β测试，其产品代号为"Tahoe（太湖）"。与Lotus不同，微软没在知识管理理论上标新立异，在这一点上，微软比Lotus"通俗"得多。

　　按照微软的说法，Tahoe是集文档管理、文档索引/检索和协同工作于一身的企业门户。Tahoe的文档管理包括版本控制、文档的作者与密码属性管理、文档发布控制、签发控制等功能。在文件索引方面，Tahoe可以进行全文检索，也可以对网站、文件系统、Exchange服务器、Lotus服务器等多种信息源进行检索。

　　除此之外，在Tahoe系统中还可以采用人工方法对文档进行分类处理，在处理过程中，Tahoe的分类助理可以学习人工分类规则，当样本达到一定数量，分类助理就可以自动进行分类。

　　Tahoe由文档服务器、索引服务器和检索服务器组成。这些服务器既可以安装在一台机器上，也可以分装在三台机器上。使用时，既可以以WWW方式进入Tahoe，也可以通过MS Office中的Tahoe插件进入，还可以直接从Windows文件系统进入。

　　微软的策略是只提供知识管理系统平台，而针对各行各业的知识管理解决方案由经微软认证的合作伙伴提供。

　　Lotus与微软在知识管理领域各有所长:Lotus像一个老谋深算的智者，微软更像一个门槛精的小伙。Lotus先建立一个严格的体系，然后再一步步推进;微软则不太在意体系，缺什么就做什么。Lotus的难点在于它所建立的体系是否能被各界人士认同，而微软的问题是体系不严密。　　IBM:挖掘文本

　　在文本挖掘软件中，IBM的Text Miner很有代表性，其主要功能是特征抽取、文档聚集、文档分类和检索。

　　Text Miner的特征抽取器能从文档中抽取人名、组织名和地名以及由多个字组成的复合词。此外，特征抽取器还能抽取表达数字的词汇，例如，"钱"、"百分比"、"时间"等。抽取完特征以后，有相似特征的文档就被自动聚集成一个集合。利用这一功能，知识管理系统可以从大量文档中找到相关文档。Text Miner还可以对文档进行自动分类。

　　Autonomy:去除冗余

　　在中国，知道Autonomy公司及其技术的人不多。但实际上，Autonomy及其CEO迈可·林奇（Mike Lynch）在知识管理界的知名度很高。迈可·林奇1991年毕业于剑桥大学，主修神经网络。他受模式识别所用的概率算法的启发，创立了Neurodynamics公司，以概率论中的贝叶斯公式和香农的信息论作为其技术的理论基础，开发出文本挖掘产品。1998年，Autonomy公司看中林奇的技术，以400万美元并购了林奇的公司，林奇也成为Autonomy公司的CEO。

　　Autonomy最核心的产品是Concept Agents。在经过训练以后，它能自动地从文本中抽取概念。

　　在林奇看来，按照香农的信息论，文档中除有效概念外，还有大量的冗余信息。而词或短语是否为冗余可根据它在文档中的随机度（概率）来判定。如果能滤去冗余，就可从文档中自动抽取出表达文档主题的概念。在林奇的方案中，先要对系统进行训练，处理一些文档，由使用者对非冗余概念做出认定和识别。按照贝叶斯概率理论，这一步实际上是让系统获得关于概念的先验概率。系统在随后的自动处理中根据这些概念在文档中出现的实际情况，按贝叶斯公式求出后验概率，以此作为冗余过滤的依据。这一方法与语种无关，由于每个用户都要对系统进行个别训练，因而系统的文本挖掘天然就具有高度个性化的特点。到目前为止，包括报业巨头默多克的新闻集团在内的一批知名公司已经成为Autonomy的客户，Compaq公司也已经将Autonomy的技术和产品纳入其知识管理解决方案并在客户中推广。

　　TelTech:服务知识管理

　　TelTech的创始人Joe Shuster是一个化学工程师，他曾创建并出售了一个成功的低温工程专业公司。这一段工作使Shuster深切感受到从公司外获取专业知识的困难。基于此，Shuster于1984年创建了TelTech公司。

　　TelTech 提供三类服务:第一类服务由专家提供。TelTech拥有数千名签约专家，他们主要是有成就的学者、退休的资深专业人士和愿意提供资询服务的专业人士。TelTech并不试图将这些人的知识存入计算机，再以专家系统的方式提供服务，而是维护专家档案,当客户需要用服务时，TelTech的知识工程师就帮助客户分析问题，并向客户推荐数位专家。第二类服务是专业文献检索，用户可以自已通过TelTech的门户网站进行检索，也可以在知识工程师的帮助下进行检索。第三类服务是产品与厂商检索，这种服务也是通过其门户网站提供。

　　TelTech成功的关键是建立了高性能的知识结构。它采用主题法，其主题词表分为不同专业，共有3万多个，由数位知识工程师维护，每周更新500～1200个词。

　　目前，计算机世界网也在致力于开发基于公共信息的知识管理系统，所采用的策略与TelTech基本相同。现在，计算机世界网"e海航标"频道提供的实际上就是基于主题法的IT知识管理服务。

【转载】What's Next? A Conversation about Web Communication with XML Pioneer, Tim Bray

杨宁 — Wed, 14 Jul 2004 08:57:00 GMT

What's Next? A Conversation about Web Communication with XML Pioneer, Tim Bray

Tim Bray

Janice J. Heiss
June 2004

Consumers find themselves with a wide array of communications options: snail mail, phone, email, instant messenger, video phone, web publishing, fax, and, increasingly, Really Simple Syndication (RSS) feeds, driven by everything from newspapers to weblogs. Everywhere, individuals and businesses are trying to make the most effective use of communications technology, while IT companies decide which technologies to bet on. In this context, XML pioneer Tim Bray joined Sun Microsystems in March 2004, where, as Director of Web Technologies in the Software CTO Office, he plans to incorporate blogging software and content syndication based on the RSS format into Sun's software line, and help set the company's direction with respect to web services and search technology.

Bray was one of the three editors of the XML 1.0 specification and the author of the first "parser" software designed to read XML documents. Prior to coming to Sun, he served as chief technology officer for the visualization software company he founded in 1999, Antarctica Systems, Inc.

We met with him recently to explore the future of RSS and web-based communication.

James Gosling recently put together from scratch an RSS feed reader called JNN, for "Juicy News Network." Available under the Berkeley Software Distribution (BSD) License, the Juicy News Network feed reader is an open source project at Sun's java.net site via JLNP (Java Network Launching Protocol). Gosling commented, "The most interesting thing is what it does to be fast at startup: all news feed reading is done by a swarm of low priority threads, one for each feed. So all feeds get fetched in parallel. This is very easy to do in Java: the threading API and networking support made it all straightforward. All the sources are there, as well as a JAR file that works on any platform, and a Mac OS X installation." Any reaction?

The interesting thing about JNN is that James, to use his words, pulled it together as "a weekend hack." This illustrates how simple and straightforward syndication technology is. In this domain, the technology is the easy part.

"RSS works well in areas where information arrives at irregular intervals, like news and publications, in which you don't want to waste time looking for information."

- Tim Bray,
Director of Web Technologies in the Software CTO Office

RSS is typically used as a web publishing tool. You have said that you would like an RSS feed to your bank account, credit card, and stock portfolio. You have compared where we are in 2004 with RSS, with 1993 when the web was taking off. Where are we headed with RSS?

There are many ways to communicate. There's email, the telephone, fax, instant messaging, and increasingly, video instant messaging. RSS is one more way to communicate, and the range of communications for which RSS is applicable is quite large. RSS works well in areas where information arrives at irregular intervals, such as news and publications, in which you don't want to waste time looking for information -- you want to be told when it shows up. So, right now, RSS has a huge sweet spot for bloggers and for news sites such as the New York Times and the BBC World Service.

But a lot of useful information, such as stock market portfolios and credit card transactions, arrives at unpredictable intervals. RSS users don't need to repeatedly visit their favorite web sites to check for updates, because when the site changes, they are notified quickly with a summary of what's new. To know when one of your investments changes substantially in price, or to be able to track debits in your bank account, is inviting.

Possibilities abound. RSS might be useful for tracking change requests during a software build or for bug tracking. So, there's a lot of information that people would like to be automatically notified about.

It's too early to know all of the RSS sweet spots, or when RSS will work better than email, instant messaging, or a phone call. We do know that RSS is going to be a major part of the communications spectrum, along with Java software.

RSS is strongly identified with blogs, but as you suggest, its uses extend far beyond blogs.

Right, blogs, almost by definition, use RSS, but there are many applications of RSS outside the blogging space. Basically, anybody who's blogging now is producing RSS. The fact that so many RSS feeds exist suggests that, of course, you would want to aggregate them. And there are people who are starting to do that. Technorati (http://www.technorati.com/) aggregates huge numbers of RSS feeds, and allows you to subscribe to the aggregates, or search them in real time. It's very different from a Google search, and a potential game changer. No one can predict where all of this is going at this point. But it's a space we're very interested in, and we want to do the right thing in.

For developers who have not adopted RSS, how can it impact the way they work, communicate, and organize information?

Any business or person providing information to others would probably benefit from RSS. Already, I have spoken with my new colleagues at Sun about providing RSS streams whenever posting new content to a portal. We're not yet talking about a product, but it could happen.

"Given the decreasing cost and increasing reliability of memory, there's a growing class of enterprise applications that could run everything out of memory."

- Tim Bray,
Director of Web Technologies in the Software CTO Office

What are the biggest misconceptions that developers and IT people have about RSS/XML syndication?

People addicted to RSS tell their friends to check it out. Often, their friends say, "I don't have time to read weblogs." Remember people in 1993 saying they did not have time to surf the Internet? This is maddening, because RSS is actually a huge time-saver. There are two misconceptions. One is that RSS is just about reading blogs, and the other is that blogs are not worth reading. Some of the most influential voices in technology and business are on weblogs.

Where do you stand on the effort to merge the RSS standards groups with the Atom group, which provides an alternative or complement to RSS?

RSS and syndication in general would benefit from having a formal standard, blessed by a standards body with a carefully reviewed specification that developers could download and use. So I'm 100% in favor of getting such a standard. I don't really care whether it's called RSS, or Atom, or RSS/Atom or whatever. There are several similar versions floating around. It's not rocket science. A solid, single, agreed-upon set of rules would be ideal.

You mentioned the wide and proliferating spectrum of communication, ranging from snail mail to video phones and so on, that individuals and businesses are trying to make the most effective use of. Do you have any insights as to where this is headed culturally or technologically?

The ground is changing beneath us. The actual costs for me to send an email, or write a blog entry, or send an instant message, or to call you on the phone are all pretty close to zero. So, cost isn't really a factor in choice of medium. Rather, it's about effectiveness. We've gone from speech, to writing, to telephones, to electronic mail. Our methods of communication are now broad. Eventually, we will determine the best use of all these options.

Java Software and .NET

You have said that the claims of pundits that .NET is a threat to Java technology's future are silly. And that .NET fails to hit the 80/20 point where you do 20% of the work and see 80% of the benefits.

Java 1.0, when it first came out, was very lean and mean. And that was excellent, because people could learn it and become proficient in it quickly. Since then, the Java language has grown into its current, sophisticated, expansive shape. It's going to be tougher for .NET to replicate that kind of successful growth.

Many of the design decisions about .NET were made before it came out. But with Java 1.0, the community collectively was able to build all the extra layers that make up what we have now.

"I think we're going to beat spam."

- Tim Bray,
Director of Web Technologies in the Software CTO Office

.NET was created by a company with a historic focus on desktop applications. My view, though somewhat controversial, is that delivering applications through web browsers versus through custom applications is much preferable. And the CIOs of the world generally agree with me about this, because maintaining desktop applications increases total cost of ownership. It's much easier to deploy, maintain, and update server-based applications and interact through a web browser. And when the web browser appeared in the mid-90's, its popularity was obvious. People migrated to the browser for just about everything in very short order. Browser-based applications are the sweet spot for the entire industry.

Cellphones and PDAs with Java applications obviously have a role to play. But for the mainstream enterprise, PC/desktop, browser-based apps are the way to go. And Microsoft is all about being a desktop company. They have a place on 95% of the world's business desktops, and they have an immense depth of experience about desktop applications and how to build and deliver them. And that shows in .NET. .NET has a huge amount of user interface machinery, which is a distraction from where the real sweet spot in business is, which is on browser-based applications.

Memory -- the New Disk

You made the intriguing comment that memory is the new disk, and disks are the new tape. You argue that enterprise applications aren't architected taking this into account, and suggest that you have ideas how they could be. Would you share some of your ideas along these lines?

Very high-performance global applications, such as Google, achieve much of their performance by running in memory, and not using the traditional disk database. Given the decreasing cost and increasing reliability of memory, there's a growing class of enterprise applications that could run everything out of memory. This is going to require different thinking, both about how we build applications and about the required infrastructure. And since I think that enterprise applications written in the Java language are going to continue to be written in the Java language, we should think carefully about the kind of infrastructure necessary to support 100% memory-resident applications. I hope to work with people at Sun on this, and perhaps build some prototypes on my own.

XML Clunkiness

You are one of the three editors of the XML 1.0 specification and the author of the first "parser" software designed to read XML documents. As you know, some developers have complained that XML is too hard. They've said it's too complex, too hard for programs to parse, too verbose, and unreadable for humans to write. They say the benefits of having everyone use XML are more than outweighed by the cost of time, training, and mistakes.

I think that, despite XML's success, the APIs available to programmers for reading, writing and manipulating it have been rather clunky. It's not surprising, given XML was only finally blessed in 1998. The APIs are getting better. Yet still, the bookkeeping and code required by XML are sources of irritation. There's lots of room for improvement.

Winning the War with Spam

What about the war against spam?

I think we're going to beat spam. (You can search for spam on my blog where I've written extensively about this.) If everyone paid a penny per email sent, the cost would be very small, but the cost to spammers would be great, and spam would stop. That's my proposal.

Eclipse.org网站中关于Eclipse 3.0 中引入的RCP的介绍

杨宁 — Wed, 14 Jul 2004 08:50:00 GMT

Rich Client Tutorial Part 1

Rich Client Tutorial Part 2

Rich Client Tutorial Part 3

另外，还有IBM采用RCP构建其产品的新闻【转载自：www.eclipsenews.com】

Staunch Eclipse supporter IBM didn't waste any time in adopting the Rich Client Platform features in the new Eclipse 3.0. Big Blue's Lotus Software brand is expected to release this month a new Workplace Client Technology and supporting collaborative applications on top of the Eclipse Rich Client Platform.

"Eclipse is an extensible integration platform," said Rick Wilson, Lotus's architect of the Workplace Client Technology. "It's a powerful model."

IBM's new Workplace Client Technology comprises three server-managed client options that administrators can assign to workers, depending on their computing needs. A Web browser provides a light client for accessing applications, a rich client provides a traditional portal environment for accessing applications, and a microenvironment provides a wireless client for accessing business applications through mobile devices.

Applications that Lotus built for its Workplace platform and also based on Eclipse, including Lotus Workplace Messaging 2.0 and Lotus Workplace Document Management 2.0, can be accessed through the Lotus client model. Those applications also should be available this month.

The model is built using the new Eclipse 3.0 runtime environment features for developing rich client platforms. "Eclipse provided the underpinnings," Wilson said, for the construction of desktop applications.

The Eclipse Foundation made changes to the 3.0 release that make it possible for developers to build applications on top of Eclipse. "We refactored the basic framework within the Eclipse platform," said Mike Milinkovich, executive director of the Eclipse Foundation. New rich client platform capability includes Eclipse's window-based workbench GUI, the dynamic plug-in functional extension mechanism, help subsystem, and update manager.

The Rich Client Platform is a new area for Eclipse, but community interest and user feedback provided the motivation for building the capability into the Eclipse integrated development environment.

Lotus employed Eclipse components like the Standard Widget Toolkit (SWT), JFace, and the workbench to compose user interfaces and create things like rich client platform views and editors in its Workplace Client Technology.

Lotus developers also leveraged the thin client frameworks that traditionally serve browser content in its rich client technology. "The idea is that Lotus Workplace rich client technology can drive the views that show up on the client page through the same model that drives the content for browsers and use a portlet to define the page," Wilson said. "Portlets can be combined with portal access control and user policy to retrieve components needed by the client to realize that page." The platform brings middleware in J2EE to the client side.

This server management component of the Workplace clients and applications also is based on the Eclipse rich client platform, and is a big key of how Lotus products add value to the Eclipse platform as a plug-in. Where the Eclipse Rich Client Platform is unmanaged, Lotus added features for a managed environment. The result is a rich client that doesn't have to roundtrip for data.

"The [Lotus] rich client platform is a browser on steroids," Wilson continued. "We have a truly unique approach doing this."

Lotus also contributed to the Eclipse Foundation's efforts by helping deliver the rich client capability within Eclipse. They helped change the runtime of Eclipse components to make the framework more rich client friendly, Wilson said.
-- Rita-Lyn Sanders, Eclipse News Senior News Editor

IBM的 alphaWorks 发布了Interoperability Tool for Eclipse and .NET WinForms

其功能介绍如下：

What is Interoperability Tool for Eclipse and .NET WinForms?

Interoperability Tool for Eclipse and .NET WinForms is a Java^TM tool that allows hosting of third-party WinForm controls in Eclipse, handling of .NET events, accessing of .NET properties, invoking of .NET methods, and instantiating of .NET objects. This tool can aid in moving to the Eclipse platform while making use of investments in .NET WinForms controls.

This tool allows the following:

hosting of WinForms controls in Eclipse
accessing of properties (getting and setting values) of WinForms controls and other common language runtime (CLR) objects from the Java side including properties of complex CLR types
addition and removal of event listeners
handling of WinForms controls events from the Java side, including standard events that even require handlers of EventHandler type and events that use specialized (that is, defined by the control developer) type for event arguments
full access to event arguments on the Java side
invocation of methods of WinForms controls and other CLR objects from the Java side, including methods that have many arguments and arguments of complex CLR types
invocation of static CLR methods from the Java side, including methods that have many arguments and arguments of complex types
construction of instances of CLR types from the Java side, including construction of objects that have many constructors, constructors with many arguments, and constructors with arguments of complex CLR types
accessing of CLR enumerations from the Java side -- both as values and as enumeration objects.

How does it work?

To host a WinForm control in Eclipse, use a Java helper class and simply specify the assembly that contains the definition of the WinForm control class (either using a strong assembly name or the path to the assembly) and the full control class name. No changes in the WinForm control are required in order to allow hosting in Eclipse.

When the control is hosted in Eclipse, Java helper classes and interfaces can be used in order to access properties of the hosted control and other .NET objects, to handle .NET events, to invoke .NET methods, to instantiate .NET objects, and so on -- all from the Java code running in Eclipse.

In this version, the .NET properties, methods, and events are accessed via string names, so it may be desirable to create wrappers for frequently accessed .NET objects in order to have strongly typed access.

Further information is available in the user manual (also included in the zip file).

【网站】The Open Source Development Environment for .NET

gsyn77 — Sat, 03 Jul 2004 23:42:00 GMT

#develop (short for SharpDevelop) is a free IDE for C# and VB.NET projects on Microsoft's .NET platform

【转载】Eclipsing .NET

gsyn77 — Sat, 03 Jul 2004 22:40:00 GMT

使用 Eclipse IDE 来开发C#

Games and Information

【转载】数据挖掘技术研究进展 作者：speedfish 日期：2004年06月14日

【转载】五大知识管理(KM)的产品与服务 作者:孙定

【转载】What's Next? A Conversation about Web Communication with XML Pioneer, Tim Bray

Java Software and .NET

Memory -- the New Disk

XML Clunkiness

Winning the War with Spam

See Also

Eclipse.org网站中关于Eclipse 3.0 中引入的RCP的介绍

Rich Client Tutorial Part 1

What is Interoperability Tool for Eclipse and .NET WinForms?

How does it work?

【网站】The Open Source Development Environment for .NET

【转载】Eclipsing .NET

【转载】数据挖掘技术研究进展作者：speedfish 日期：2004年06月14日

【转载】五大知识管理(KM)的产品与服务作者:孙定