- you have to care about incompatibilities. Frameworks like Dojo do it for you, but I compare it to the portability of Hibernate/CMP 2.0/3 code between different databases. Such abstractions are not always totally transparent, in most cases you have still to now, which browser you are using.
- the idea of http and browser, was synchronous request and response at one page. Ajax and other technologies break this rule - which is an obvious hack. Nothing against hacking, but every hack needs some workaround, which are expensive to implement and even more expensive to maintain (hacks are hard to document :-))
- the communication between browser and server is more fine grained. It is hard to say, whether this fact will have impact to the scaleability and performance, but the developer has to consider this in the architecture and design. In slow networks, fine grained communication will hit the performance (because of latency).
- You can build very cool webapps WITHOUT installation of additional runtime.
It depends on the project context and expecially the customer, whether an additional installation is appropriate, or not. For the most intranet applications, the installation of an additional runtime shouldn't be a big issue.
The blog shows also a (clumpsy) work-around (http://blogs.sun.com/roller/page/sundararajan?entry=jsr_223_script_engine_for)
Posted by Carsten on July 20, 2006 at 10:05 PM CEST #
All JSR-223 languages can be compiled into bytecode.
Compilation to bytecode is only possible in a JVM. In browser is this feature not so interesting :-). Redundancy is a general problem. This can be solved either with metadata (and codegeneration) or multiple deployment of the same code. Nothing else than a hack :-)
Posted by Adam Bien on July 21, 2006 at 10:47 AM CEST #