Beyond an SOA IDE: end-to-end optimization of mashup delivery

Definitely there is an increasing need and fit for mashup platforms inside corporate IT. At OutSystems (www.outsystems.com), we have been tracking the way our customers have been building, deploying and large portfolios of mashups inside corporations for 5.5 years. I thought we might share some of our findings over the years and the challenges we have had to face as we upgrade our product.

From our experience, the key aspects of a successful mashup platform are:

• Accelerate Application Creation with Integrated Visual Modeling – The SOA IDE must include all tools to build complete applications around services, including: user interface navigation modeling, business process modeling, web services exposure/consumption, business rule modeling and persistent database modeling;
• Automate Application Deployment – Deploying and running applications defined as visual models must be an easy, error-free and fully automated activity;
• Enable Rapid Modification and Easy Adaptation to Business Changes – It must be easy to modify the components of an application, and then it must also be easy to apply and reflect those changes to all dependable components and re-deploy them. e.g. when changing the interface of a Web Service, it must be easy to reflect those changes to the related end-user interface components or business process flows;
• Foster End-user Innovation with Drag’n’Drop Composition at the UI level – It must be easy to compose the user interface content in front of a business user and it must be feasible to test it with real data;
• Managed, Controlled and Collaborative – As SOA becomes mainstream inside organizations, a lot of collaboration and control is required to enable the required degree of flexibility and agility. All aspects of traditional application development that go beyond the development tools must be integrated in the Application Delivery process. Those include, at least: integrating a services repository for future reuse; a version control sub-system; tools for reviewing and merging differences among visual models; and distributed (cross-service) debugging tools;
• Non-proprietary, Standard and Scalable – Productivity and SOA-enablement often come at a price. Since ever, when the mashups are initially built as “edge applications” our customers have required the use of standard runtime technologies (like .NET and Java), openness of generated code and attention to scalability issues. At the end, the applications are expected to perform as well as if they were written by a skilled java programmer.

I guess the list above comes close to what our customers have been telling us of what a dream mashup platform should be!

Best regards,
Carlos Alves
OutSystems (www.outsystems.com)