The small issue cited by Bob has further ramifications. Actually the geometric realization functor "should" preserve all finite limits. That is probably a main reason (along with the lack of function spaces) for the gradual demise of Top (and its relatives like Locales) as the default model of cohesion. A functor with a right adjoint and preserving finite limits is the inverse image of a geometric morphism of toposes, provided it is between toposes. Can Top be reasonably replaced by a topos that will serve all purposes of algebraic topology, functional analysis, etc ? Yes, as Peter Johnstone showed some years ago, using in fact a well-known monoid as site. The issue for simplicial realization in particular is whether - the unit interval is totally ordered or not and whether -its endpoints are distinct. That is because simplicial sets is precisely the classifying topos for such structures, as was pointed out by Joyal and explained well both in Johnstone and in Mac Lane & Moerdijk. (The second condition justifies the omission of 0 from the Delta site). Johnstone used the monoid of continuous endomaps of the generic convergent sequence to achieve the total order of the unit interval; the internal meaning of the latter is that inside the square there is no subobject containing both solid triangles. No sheaf, that is. But Peter achieved that solution after detailed study led him to reject another model, proposed by several people whose geometric intuition does not include Peano curves, undecidable statements, etc (e.g., me). That proposal, namely that the basic figures of topology are continuous curves, had to be rejected at the then-current level of knowledge because the usual model for the shape of these figures, the unit interval as constructed in traditional set theory, admits far too many endomaps, along which coverings must be stable by pullback; thus too few coverings , too many sheaves. This leads to the reasonable demand for a submonoid of those continuous reparametrizers, containing polynomials and lattice operations but not containing, for example, the coordinates of a Peano curve in the square. That demand is very similar to the one put forth by Grothendieck in his proposal for TAME TOPOLOGY. I pointed out several years ago that Grothendieck's demand is related to the achievements of mathematicians working on so-called O-minimal models, and some of the of them in fact mention Grothendieck's slogan in their discussions. But I am not aware of any publications addressing the issue exposed by Peter. Can we now understand by example the kind of monoid desired ? Best wishes Bill On Sat 04/25/09 2:54 AM , Robert L Knighten RLK@knighten.org sent:
Dear Hugo,
Your question involves the functors
N | | Cat -----> SSet ------> Top
(nerve and geometric realization) and their composite, the classifying> space functor B.
1. The nerve functor N has a left adjoint, so in particular it
Tom Leinster writes: preserves> finite products. Hence if M is a monoid in Cat (i.e. a strict monoidal> category) then N(M) is, in a natural way, a monoid in SSet.
2. It's also true, though not totally obvious, that the geometric realization functor | | preserves finite products. So if X is a
monoid> in SSet then |X| is a topological monoid.
A small issue - this works provided the destination of the geometric realization is the category of compactly generated Hausdorff spaces. Otherwise, as in Milnor's original paper, there are limitation on the simplicial sets involved.
-- Bob
-- Robert L. Knighten RLK@knighten .org