Many approaches to software architecture assume that architecture be planned at the beginning from the project's quality goals. This is problematic as the macroarchitecture is hard to change, and the quality goals it's based on tend to be unknown at the beginning of a project, or change later. Consequently, it would really be preferable if we could defer the macroarchitectural decisions until later.

This talk shows how to do this using systematic modelling and functional programming.

Target Audience: Developers, Architects
Prerequisites: Basic architecture knowledge
Level: Advanced

Extended Abstract:
What should we do first then? We should write down what we know, at the time we know it, and do it in such a way that we generate reusable components that can be assembled into any macroarchitecture.

Concretely, this can be done by using staple techniques from functional programming:
•    building small, flexible combinators instead of fixed attribute-structure OO models
•    decoupling pervasively using abstraction
•    using immutability to avoid hidden dependencies

The talk shows how to do this, and report on our experience in several concrete projects.