Best of this Week Summary 20 January - 25 January 2009

• Five performance anti-patterns in the data-access layer in database-driven applications. The following are described:
  
  
  • Misuse of O/R Mappers: loading behaviour and loading time (e.g lazy loading)
  
  
  • Load More Data Then Needed
  
  
  • Inadequate Usage of Resources: database connections
  
  
  • One Bunch of Everything: all data access is treated equally (e.g access frequency)
  
  
  • Bad testing


• An investigation on how the front-end of iWork.com has been built. Used are SproutCore(Javascript framework), Prototype, WebDAV and JSON.
  


• 50 minutes video presentation on Orbitz.com's architecture. Brian Zimmer describes a search and booking example for seats and beds. Problems they came across for example are: i18n, distance estimations. Disclaimer: the architecture is not perfect and always in motion. Orbitz.com is a Java shop, amongst others they use: Spring, Spring Web Flow, Tomcat and Joda Time (side-step: much learned from the Joda Time project is being implemented in JSR 310, led by the Joda Time project lead), Log4J, searchable (query-like) complex event processing mainly for logging/monitoring(!) and throttling, multi-threading, caching, availability and Jini.
  


• Performance comparison of four JPA implementations: Hibernate, Toplink Essentials, OpenJPA and EclipseLink. Only JPA was used in the tests. Conclusion: there's no clear winner, all have good and bad things on different points like CPU or memory.
  


• Will building desktop Linux applications with Javascript be the next re/evolution for Javascript?