I’ve been exploring using Flex with Salesforce, using the Flex Toolkit as a connector to the database. Server side functionality over and above persistence is implemented with their Apex code stored procedures platform. The advantage is I can deliver SaaS products very quickly – no need to manage a server farm.

The concerns I’m seeing here is with security. For example, if I use a role-based permissions framework to expose appropriate functionality on the client, I’m feeling exposed to someone messing with returned roles from the server, etc.

Any thoughts on the security implications between these two approaches?

Now we’re trying to make simple but reusable CMS, that shoul be configurable (without writing any code) for any database model. This means we have to make client-server architecure for passing any data – flex-java class mapping is impossible, because we don’t know data model while wrting the framework. The idea is to create something like queries for objects defining fields, pagination, filters etc.

I think the choice between client-centered or server-centered architecture should be made in context of the problem to solve.

we are working on Flex based assessment tools that have very big domain models. The role of the server is in our case mainly reduced to CRUD operations. We use Cairngorm and find it to become more difficult to expand on what we are doing because everything is so coupled. To prevent this, we’re using the Prana IoC container (of which I’m the author). It solves some of the dependency lock-ins you have with Cairngorm, but I still feel like we too need another architectural framework for our kind of application. That’s why I’m thinking about adding an MVC implementation in Prana that focuses on the domain model. If you are interested in sharing ideas or co-operating, please contact me.