New CA CEO is Bill McCracken
CA Have announced their new CEO – Bill McCracken (here: http://www.ca.com/us/press/release.aspx?cid=227312) – read about his background here hopefully he will be as supportive of CA Gen as John Swainson was and will help the product and the community move forward!
The limitations in web service access is given in fact by the technology used. In the web environment where you reside on a web server the mechanisms of reaching other web components are easy. But a host Tx running for example in CICS cannot access web services that easily because of two reasons . The one is speed and waitig time till timeout. One milisecond in an host environment is too much time and the running transactions would get some synchronization problems, when they had to wait for program responses from web.. One possible way could for example be the asynchron usage of MQ series, to access web services from a host transaction. Another problem is the demand of beeing neutral (platform independend) for an actually platform independend pseudocode. Ported to windows environment those external service calls may not work as well. One has to build some sort of (reverse -joke-) proxy for web access. Regarding the storage of additional metamodel elements again the bottleneck of reinventing everything from the scratch is the show stopper. Why not to make the own metamodel extensions in eclipse and synchronize with CA Gen repository using some access classes as ProgGen does. Eclipse environment offers generic MM extension capability. I think the real issue is also here a seamless tool integration .. Each tool should concentrate on its own strengths…
I agree with Danny – but specifically on the Web Service Access Designer, it only allows access to web services from a web page – it doesn’t allow access to web services from other types of technology.
What I’m driving at is that many web services are consumed by, for example, C -generated servers on a Unix platform. After all, isn’t the whole point of standards-based service-access the fact that any technology can be accessed by any other technology via a standard interface (viz: a WSDL definition) ?
Concerning storage of things in the model, it would be of extreme use here if we could somehow link the model to a UDDI service, or some other service discovery infrastructure, so that if a remote service changed definition, we could impact assess that change and determine which portions of our model would be affected and need changing.
Although I have no authority what so ever to comment on what Bill will be supportive of or not, I’m thinking there will not be that much of a change here, so we’re good…
I would like to make the comment though that we also need the community to help the product move forward!
The power of model-based development is that higher-level more “abstract” platform-neutral modeling artifacts eventually drive the generation of code; we call that the “downstream effect”. Not only does this give us productivity, i.e. it allows us to concentrate on defining the business logic as opposed to loosing our way in the technical complexity of various low-level APIs. It also give us governance; we list some of the higher-level modeling artifacts and we know that this is what is implemented and nothing more or less. E.g. if we got a “sensitive” attribute in some table and some compliance rule demands us to know where it is used, then we know our “where used” report is correct and we don’t need to do anything more, we do not have to wade through hundreds of source codes checking whether or not something in the code touches it.
So where does the community need to help? Well, you need to tell the development organization what are the high-level platform neutral modeling artifacts that you need in order to drive your “down stream” effects, i.e. drive your productivity, and to give you governance.
The latest Gen releases have focussed a lot on the lower-level technical capabilities. Take Gen r8 as an example, there are some very exiting features like Web Service Access Designer and Procedure Step Interface Designer. But it is unfortunate, in my view, that when you are using these features there is very little information which is actually added to your Gen model.
Web Service Access Designer offers an easy way to invoke Web Services from within HTML pages, however which web services you are invoking isn’t stored in the Gen model. So, from a business point of view you can have a use case, i.e. could be a Gen generated WebView page, where the realization of that use case is not completely realized by Gen business logic itself but also by an additional Web Service that you invoke. That fact which can be important for governance can not anymore be retrieved from Gen model information.
Similar issues for Procedure Step Designer, here you can easily add additional operations starting from one basic operation, i.e. a Procedure Step. Let’s say you have a multi-function procedure step where depending on the COMMAND value, e.g. CREATE, UPDATE, DELETE, different business logic is executed. You can now very easily extend that interface by adding a Create, an Update and a Delete operation. But if this original procedure step was an operation of an InterfaceType in your CBD architecture, then this interface now in reality has 3 additional operations. But the CBD diagrams don’t show this fact. So, how now to properly assess if an Interface is suited for some new development project?
Anyways, just some examples of where the user community needs to be vigilant and tell us what are the higher-level platform-neutral modeling objects you need in order to be even more productive and to ensure your governance.
Cheers,
Danny