2-filteredness and the point of every Galois topos (Q969693)

The notion of Galois object in a pointed topos can be thought of as related to the notion of normal extension (`Artin's interpretation'). Dubuc's paper gives a definition of Galois object in a bounded topos which can be made even when the topos has no point. A bounded connected locally connected topos is then called Galois if every connected object is covered by a Galois object. The paper proves that, in fact, every Galois topos has a point. The point is constructed by relying on a characterization of any Galois topos as a 2-cofiltered bilimit of toposes. Even though a Galois topos, by definition, can be seen as a filtered union of sites, each with a point, the points do not necessarily form a cone. To overcome this obstruction the paper provides an alternative characterization of these sites such that they form the objects of a 2-filtered system for which the points do form a cone and so give rise to a point of the 2-cofiltered bilimit topos. The paper omits some detail on how the 2-cofiltered bilimit topos obtained from this alternative characterization of the sites, is equivalent to the original bounded topos and so is not entirely self-contained. However further details on this point can be obtained from a later paper by \textit{E. J. Dubuc} and \textit{S. Yuhjtman} (`A construction of 2-cofiltered bilimits of topoi') which is now on arXiv (2011; 1107.1685v1).