All topoi are localic or why permutation models prevail (Q1087632)

The appearance in print of a paper by Freyd is a rare event; and category-theorists therefore have cause to rejoice that this one, which was written in 1979, has at last found its way into the pages of J.P.A.A. The problem it addresses is the following: can every (Grothendieck) topos be built up from the topos of sets by means of the two constructions of (1) forming the topos of sheaves on a complete Heyting algebra and (2) forming the topos of continuous G-sets for a topological group G ? Freyd shows that if we add (3) cutting down to exponential varieties (by which he means full subcategories closed under arbitrary products, subobjects and power-objects), we can indeed get every Grothendieck topos; moreover, we need only one instance of (2), corresponding to a group G which may be either the group \(G_ 0\) of all permutations of \({\mathbb{N}}\), or the group \(G_ 1\) of order-preserving permutations of \({\mathbb{Q}}\) (with the topology of pointwise convergence in each case), or a dense subgroup of either of these. He also provides a simple example to show that (1) and (2) do not suffice. The paper is published almost exactly as originally written, with no references to subsequent work (and indeed no formal references of any kind). It should perhaps be mentioned that the reviewer [Math. Proc. Camb. Philos. Soc. 93, 409-419 (1983; Zbl 0515.18002)] has characterized those toposes which are obtainable using (1) and (2). Once again, \(G_ 0\) suffices to reach all of these, but \(G_ 1\) does not, even for the Boolean ones - the latter fact follows from work of \textit{G. P. Monro} [J. Symb. Logic 48, 39-52 (1983; Zbl 0522.03034)].