CADE - 21 The 21 st Conference on Automated Deduction

Many years ago various new concepts have been introduced into Isabelle to support the Isar proof language. While the original motivation was coming from the demands of human-readable proof texts, most of this infrastructure has turned out as generally useful in building other advanced mechanisms. So today “Isar” concepts already permeate large parts of Isabelle, enabling simpler and more powerful implementations. Here the central principle is that of a local proof context, which allows to work relative to fixed declarations, producing exported results eventually. The structure of a proof context is motivated by the Pure logical framework (with type variables, term variables, assumptions), but is not limited to that: context content can be arbitrary data represented as ML types internally. This allows all kinds of derived mechanisms to coexist within the same local environment. Instead of manipulating bare-bones theorem and theory values (according to the original “LCF approach”), one may work within a rich environment in a structured manner. There are extra possibilities to transfer results into different application domains by re-interpreting portions of the context later on. This talk provides an overview of these new possibilities from the user-perspective, covering the impact on proof texts, theory specifications, and everything. At the same time, it gives an impression of what the forthcoming Isabelle2007 release will have in store. 1 Isabelle’07, page: 1