The Paris JUG this time was devoted to GWT and implementation of Restlet for GWT. Two presentations greatly performed by Didier Girard and Jerome Louvel. I was impressed by the number of participants, thinking that GWT was already in the soul. Bravo to Didier continue to promote this technology and Jerome to put the concept of REST.
Still, I am astonished at the vision that people have of the RIA. For many I have the impression that RIA means rich web application. But the notion of « rich » is simply a reconsideration of the user experience and this does not necessarily by the browser. GWT, Ajax, AppletFX, Flex or Silverlight do not have the monopoly of RIA simply because they run in a browser.
iTunes is probably one of the first RIA and this is not a web application. As I said already we must differentiate Rich Web Application and Rich Desktop Application. A RWA runs in the browser, a RDA runs in an
There are advantages and disadvantages in both approaches but the distribution and maintenance which have long been the strengths of Web applications are no longer true today.
As you do more surprised to have your Windows automatically updated it is now also possible for an RDA to be maintained and transmitted via the Web. Java Web Start is a good example and Adobe AIR knows how to do too. But i herer what you say : to run Java or Air, you have to respectively install the JVM and runtime AIR. Yes it’s true like you do it recently when you installed Chrome or previously installed Firefox. The difference is that Windows come with its own browser and when you buy a PC with Windows you have it. At one time you install a runtime and except if you are terribly lazy person you aren’t obliged to be use Windows/IE only.
Where we must be vigilant with a RDA is to not go back to the old client/server model. The RDA must remain a UI layer and not shipping business that must always be server side.<br /> Apple had already experimented this with WebObjects Java Client and continues today to offer similar APIs for Cocoa and now IPhone SDK.
To the slogan « the browser is the platform » I answer « the
browser is a platform ». Future architectures should not be dependent on UI layer because it must be developed in the best technology to meet the objectives of application and customer needs by focusing on ergonomics and performance.
Our application ResUrgences is in web for 8 years now in a sector (health) where applications are often client/server. Now the extension from the emergency department to the mobile emergency department (say SMUR in France) forces us to reconsider the web architecture because using a web application on a Tablet PC offline, although not impossible, is not appropriate. Especially when you have to interconnect with monitoring and electro-cardiogram equipment.
So what choice between RWA and RDA ? The first step is to think RIA, so to have a SOA server side architecture and services accessible through various formats via various protocols. At this point different criteria have to be focus on: ergonomics, performance, accessibility, environment (OS and hardware), openness, security, compatibility with existing … So there is no obvious answer. I’m looking for a long time on a table that defines what technology for what needs and now I think it’s impossible.
What should be considered is :
- access to local resources (files, applications, USB, serial, Bluetooth) is an argument to look to the RDA. Although this can be remedied with an applet and increasingly with Gears (but it’s a RWA/RDA mix, why not is what I do) and that the Flash plugin already allows access to the camera.
- the concept of « push » to send data to a client from the server is now possible with RWA (Comet, reverse Ajax) and soon standarize in HTML 5 (WebSocket).
- exchanges via asynchronous MOM with queues client side cannot yet be in RWA. Gears should propose an API and LifeCycle Data Service does not really do it because the queue is server side.
- uniformity of application with a single platform deployment independent of the OS is probably the most powerful argument for the SI to choose a RWA