On 4/11/07, Jean-Sebastien Delfino <[EMAIL PROTECTED]> wrote:
Raymond Feng wrote: > Hi, > > I think we should start to leverage the extensibility story (see the > thread titled "The pluggability for Tuscany runtime extensions") > instead of hard-coding all the processors into the SimpleRuntimeImpl. > > Thanks, > Raymond > > ----- Original Message ----- From: <[EMAIL PROTECTED]> > To: <tuscany-commits@ws.apache.org> > Sent: Wednesday, April 11, 2007 10:14 AM > Subject: svn commit: r527581 - in > /incubator/tuscany/java/sca/modules/host-embedded: pom.xml > src/main/java/org/apache/tuscany/host/embedded/SimpleRuntimeImpl.java > > Raymond, I agree that we need to leverage the runtime extensibility story, for runtime extensions :) But are the WSDL and Java interface support really extensions or part of the base SCA assembly support? On one hand, no, they are not real extensions as they are part of the base SCA assembly spec. So we don't need to handle them as extensions, SimpleRuntimeImpl can depend on them the same way as it depends on the assembly-xml module. On the other hand, we could consider them as extensions, as Java and WSDL are just two of many programming languages that can be used to define service interfaces. To do this we just need two new Maven modules hosting their ModuleActivators. Going further, we could even consider assembly-xml as an extension, and have a ModuleActivator for it as well. So, our new modularized kernel structure allows for both approaches. It's a matter of where we want to draw the line between core SCA and extensions. At the moment, I'm tempted to follow the structure of the SCA specifications to draw these lines. What's in the SCA assembly spec is part of the base, what's not in the SCA assembly spec is extensions. But I could be convinced otherwise... Thoughts?
I'd be leaning to making them proper extensions unless it makes the code significantly more complicated. One of the problems we've had in the past is that by not making things like Java 1st class extensions then when it comes to adding something else as a real extension various issues get exposed. Another thing is that the SCA assembly spec also says that the WS binding MUST be supported by an SCA runtime so following the argument above our Axis2 binding could be hardcoded into the runtime which doesn't seem right. ...ant
