In T 570/91 the board emphasised that although a person skilled in the art was completely free in choosing a starting point, he would of course be bound afterwards by that choice. If, for instance, the skilled person preferred and decided to start from a specific compressor piston, he could further develop that piston but at the end of that development the normal result would still be a compressor piston and not an internal combustion engine piston. In T 439/92 it was explained that a conscious choice of starting point, made in the knowledge of the respective benefits and drawbacks of the various types concerned, not only determined the subject-matter serving as a starting point but also defined the framework for further development, i.e. a further development within this particular type. A change of type during the further development of the consciously chosen type, to another type, which was previously known but had not been chosen, could then only be seen as the result of an expostfacto analysis (see also T 1040/93, T 35/95, T 739/95, T 255/03). It is unlikely, and normally not obvious, for the invention type originally chosen to be changed during development (T 817/94). A generically different document cannot normally be considered as a realistic starting point for the assessment of inventive step (T 870/96, T 1105/92, T 464/98).
Following T 439/92 the board stated in T 1228/08, that although D1 disclosed the use of hierarchical coding, which was a major aspect of the present invention, the board considered that this document did not represent a realistic starting point for an attack on inventive step. A conscious choice of a starting point not only determines the subject-matter serving as a starting point but also defines the framework for further development. Thus, using D1 as a starting point would mean that any further development would be carried out in the context of multicasting: it is unrealistic to suggest that, starting from this disclosure, the skilled person would go outside this framework - indeed take a technical step back - to develop a non-multicasting system.
In T 487/95 the board chose a military protective helmet as the closest prior art, pointing out, however, that this did not mean that documents describing protective helmets of a different kind (such as workers' safety helmets) could not be said to form part of the knowledge of a person skilled in the art. In this case, adopting the problem and solution approach, the information contained in the patent which related to a known military protective helmet (D9) represented the primary source of information, i.e. the most promising point of departure, from which the skilled person would attempt to arrive at the claimed subject-matter. The other documents could, however, represent important secondary sources of information (in this case: workers' safety helmets) from which the skilled practitioner could obtain indications and suggestions with regard to the problem being addressed (see also T 149/00).