Device Independence for GUI, Voice and Multimodal Application: is it possible ?

Is the single authoring concept of DI applicable for non-GUI modalities ?
Can we develop modality-independent authoring techniques/technologies which would render gracefully in any modality ?
What are the particular challenges for rendering an application in a voice-only environment?
Are voice interactions very different from GUI interactions making voice applications a challenge for DI ?

Topic 3 : Vision