REVIEW 2 major objections 4 minor 6 references
Local fibrations and morphisms of relative toposes
T0 review · 2 major / 4 minor · reviewed 2026-08-06 · deepseek-v4-flash
Pith's one-line read A site morphism induces a relative topos exactly when it preserves locally cartesian arrows.
desk verdict Genuinely new notion and a clean criterion, but the proof of the main sufficiency has a gap where a stack-level colimit-cofinality lemma is being invoked as a site-level fact. read the letter →
The pith
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
The reading
What carries the argument
The central objects are locally cartesian arrows and local fibrations. An arrow f in a comorphism of sites is locally cartesian when η(f) is cartesian in the canonical relative site, which concretely means it admits liftings and uniqueness up to a covering; a local fibration is a comorphism in which every arrow into the base can be locally factorized through locally cartesian arrows. The proof of the main theorem transfers cartesian arrows in the canonical stack to the site by two localizations: first by pullbacks along covering families, then by colimit presentations, and the η-extension ~A* is the functor that carries this local data to the target relative topos.
What would settle it
In the canonical stack of a topos with binary coproducts, the arrow $(\nabla_c,1_c):(c\sqcup c,c,\nabla_c)\to(c,c,1_c)$ is not cartesian, although both coprojections make it cartesian; Proposition 4.2.6 must recover this failure from the cofinality condition on $D^{\eta-\mathrm{fact}}_{(f,g),d}$. Verifying that this single arrow is not locally cartesian is a concrete check of the colimit-localization that Theorem 4.3.1 depends on.
Extended reading notes
Core claim
The core discovery is that the right topology-aware relaxation of 'fibration' is obtained via the canonical stack: an arrow is locally cartesian when its image under the canonical functor η is cartesian, and a local fibration is a comorphism of sites in which every arrow into the base has local liftings by locally cartesian arrows. Theorem 4.3.1 states that for a local fibration p, a (J,K′)-fibration p′, and a morphism of sites A with p′A ≃ p, the induced geometric morphism Sh(A) is a morphism of relative toposes iff the η-extension ~A* is a morphism of fibrations iff A is a morphism of local fibrations. This gives a necessary and sufficient site-level condition for relative morphisms, and it specializes to a relative Diaconescu theorem for local fibrations.
Load-bearing premise
The proof rests on the claim that in the canonical relative site of a geometric morphism, every locally cartesian arrow is already cartesian, so that local liftings over covering families can be glued into global liftings; if that gluing fails, the whole criterion collapses.
Editorial extensions
If this is right
- A morphism of sites between a local fibration and a fibration over the same base induces a relative geometric morphism exactly when it is a morphism of local fibrations.
- The relative Diaconescu equivalence extends to local fibrations: K-flat functors from a local fibration to a relative topos correspond to morphisms of local fibrations that are also morphisms of sites.
- A continuous local fibration induces a locally connected geometric morphism, and an essential geometric morphism is locally connected if and only if its essential image is a fibration.
- Totally connected geometric morphisms are exactly essential morphisms whose essential image is a cartesian fibration, and the canonical stack construction provides the free totally connected relative topos.
- Weak indexed geometric morphisms are equivalent to continuous morphisms of local fibrations, and the Giraud topos of a fibration is canonically equivalent to that of its stackification.
Reading between the lines
- The criterion suggests a practical way to check whether a given site morphism extends to a relative geometric morphism: instead of global cartesianness, verify the local lifting property, which is a finite condition when coverings are manageable.
- For ordinary fibrations, the difference between being a morphism of fibrations and a morphism of local fibrations is invisible at the topos level, so sheafification can repair failures to preserve cartesian arrows.
- The fibrancy characterization of locally connected morphisms could serve as a definition in settings where the essential image is available but the inverse image functor is hard to compute.
- The weak indexed Diaconescu equivalence suggests that continuous morphisms of local fibrations are the natural site-level shadow of weak geometric morphisms generally, which may clarify base-change questions.
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The paper introduces locally cartesian arrows and local fibrations, a topological refinement of classical fibrations, and uses them to characterize when a morphism of sites between relative sites induces a morphism of relative toposes (Theorem 4.3.1). It also proves a relative Diaconescu theorem for local fibrations, gives fibrational characterizations of locally connected and totally connected geometric morphisms, and proves an indexed weak Diaconescu theorem. The paper is written in the framework of relative topos theory via stacks and relies substantially on the authors' earlier work [BC23], [CZ21], and [Car20].
Significance. If the main results are correct, the paper provides a genuinely useful refinement of fibration theory: local fibrations sit between ordinary fibrations and arbitrary comorphisms of sites, and the criterion of Theorem 4.3.1 gives a complete site-level characterization of morphisms of relative toposes. The applications to locally connected and totally connected morphisms (Propositions 5.2.8 and 5.4.1) and the indexed weak Diaconescu theorem (Theorem 6.3.3) are attractive and would be of independent interest. The exposition is ambitious and contains many precise statements, and several definitions are worked out in detail. However, the central equivalence has a proof gap in the sufficiency direction, and the paper also contains several deferred proofs and missing references that need attention before the main claim is fully supported.
major comments (2)
- [Section 4.2, Proposition 4.2.8] The step labelled "Thus, we are in the situation of (iii) 4.2.8" is not justified. Proposition 4.2.8 is a criterion for an arrow f : d' -> d in the site D, whereas the proof at that point concerns an arrow (f',g') : P -> eta_D(d) in the canonical stack (Dhat_K/C*_p). The needed statement is a stack-level analogue: an arrow with codomain eta_D(d) whose domain is presented as a J-cofinal colimit of arrows eta_D(f_i) with f_i locally cartesian should be cartesian. Proposition 4.2.6 is only a partial step toward such a lemma, and it requires the whole category D^{eta-fact}_{(f',g'),d} to be cofinal; the proof does not show that this cofinality is inherited by the image category after applying ~A*. As written, therefore, the implication (iii) => (i) is unsupported. This is load-bearing for the main theorem and must be repaired by stating and proving the missing stack-level colimit criterion.
- [Section 4.2, Proposition 4.2.8] The statement of the cofinality criterion is ambiguous and possibly incorrect as written. In the displayed diagram, the functor pi_C^f has codomain C, whereas the proof of (iv) and the intended use in Theorem 4.3.1 suggest that the relevant functor should land in C/p(d), sending (d'',v'',f'') to (p(d''), p(f'')). The claim in the proof that (i) implies (ii) is "immediate" because (d',1,f) is an object of D^{cart-fact}_f is not enough for cofinality into C: for a general object c of C, the comma category (c/pi_C^f) need not be nonempty. If the intended notion is J-cofinality of the functor D^{cart-fact}_f -> C/p(d), the statement and proof should say so explicitly. Because Proposition 4.2.8 is invoked in the proof of the main theorem, this ambiguity affects the central argument.
minor comments (4)
- [Section 2.4] Both propositions contain the placeholder marker "[?]" in place of a reference ("as already expressed in [?]" and "as it can be seen for example in [?]"). These citations should be supplied.
- [Section 4.2, Proposition 4.2.1] The notation for the covering arrows is inconsistent: the statement says "cartesian arrows ((v_i,u_i) : eta_D(d_i) -> [alpha])_i" but then writes "(f,g) ∘ (u_i,v_i) = eta_D(f'_i)". The order of the two components should be fixed throughout.
- [Theorem 4.3.1] The expression "(J,K')-fibration" is used in the statement without a definition. The authors should clarify whether this means a fibration whose topology K' contains the Giraud topology for J, and how it relates to the definition of relative site in Definition 2.1.8.
- [Various proofs] Several technical claims are deferred with "lengthy but straightforward" (Proposition 2.4.8, part of Proposition 2.4.15, part of the proof of Proposition 5.2.6, and the naturality part of Proposition 5.4.6). Given that Section 2.4 underpins the view of the canonical stack as a topos, at least the most central of these proofs should be expanded or replaced by a precise reference.
Circularity Check
No circular reduction found; the paper's new equivalence is not forced by its definitions, though it leans on self-cited prior results and contains a non-circular proof gap.
full rationale
After walking the derivation chain, I find no step in which a claimed prediction or theorem is equivalent to its inputs by construction. Locally cartesian arrows are defined via the canonical functor eta_D (Definition 3.1.1), and morphisms of local fibrations are defined as preservation of these arrows (Definition 3.2.5), so the necessity direction (ii)=> (iii) of Theorem 4.3.1 is indeed immediate from the definition, as the paper openly says. The sufficiency direction (iii)=> (ii) is not a tautology: it uses the commuting square ~A* eta_D = eta_D' A (Proposition 2.3.1(5)) and the two localizations of cartesian arrows in Sections 4.1-4.2; if the argument were valid, it would be a substantive theorem rather than a renaming. The equivalence (i)<=> (ii) is imported from the authors' earlier work [BC23], which is a self-citation, but citing one's own prior theorem is not itself circular; the new condition (iii) is not defined as (i). The proof of (iii)=> (ii) does contain an apparent gap: after presenting ~A*(P) as a colimit of eta_D'(A(d_i)), the text invokes 'the situation of (iii) 4.2.8', but Proposition 4.2.8 is a site-level criterion about arrows of D, not a stack-level criterion about arrows of the canonical stack with such a colimit presentation; the needed cofinality of the relevant comma category is not verified. This is a correctness risk, not a circularity: no equation in the paper makes the conclusion equal to a hypothesis. The score of 2 reflects only the repeated load-bearing self-citations ([BC23], [CZ21], [Car20]) in the setup, not a circular reduction.
Assumptions & free parameters
assumptions (4)
- standard math Giraud's characterization of Grothendieck toposes
- standard math The equivalence between fibrations and indexed categories, including the pullback and two-out-of-three properties of cartesian arrows
- domain assumption Prior relative topos theory results from [BC23], [CZ21], [Car20]
- domain assumption The base topos has finite limits and its canonical topology makes covering families effective-epimorphic
Cite this review
Pith. "Pith review of Local fibrations and morphisms of relative toposes." pith.science (2026). https://pith.science/paper/5TNAM7CN
@misc{pith2026250714379,
author = {Pith},
title = {Pith review of: Local fibrations and morphisms of relative toposes},
year = {2026},
howpublished = {\url{https://pith.science/paper/5TNAM7CN}},
note = {Machine review of arXiv:2507.14379}
}
read the original abstract
We introduce the notion of local fibration, a generalization of the notion of fibration which takes into account the presence of Grothendieck topologies on the two categories, and show that the classical results about fibrations lift to this more general setting. As an application of this notion, we obtain a characterization of the functors between relative sites that induce a morphism between the corresponding relative toposes: these are exactly the morphisms of sites which are morphisms of local fibrations. Also, we prove a weak version of Diaconescu's theorem, providing an equivalence between the continuous morphisms of local fibrations towards the canonical stack of a relative topos and the weak morphisms between the associated indexed toposes. The paper also contains a number of results of independent interest on morphisms of toposes and their associated stacks, including a fibrational characterization of locally connected (resp. totally connected) morphisms.
Reference graph
Works this paper leans on
-
[1]
On morphisms of relative toposes, 2023
Léo Bartoli and Olivia Caramello. On morphisms of relative toposes, 2023
work page 2023
-
[2]
Denseness conditions, morphisms and equivalences of toposes
Olivia Caramello. Denseness conditions, morphisms and equivalences of toposes. arXiv:math.CT/1906.08737v3, 2020
arXiv 1906
-
[3]
Fibred sites and existential toposes
Olivia Caramello. Fibred sites and existential toposes. arXiv:math.AG/2212.11693, 2022
arXiv 2022
-
[4]
Relative topos theory via stacks
Olivia Caramello and Riccardo Zanfa. Relative topos theory via stacks. arXiv:math.AG/2107.04417, 2021
arXiv 2021
-
[5]
Jean Giraud. Classifying topos. In F. W. Lawvere, editor, Toposes, Algebraic Geometry and Logic , pages 43--56. Springer Berlin Heidelberg, 1972
work page 1972
- [6]
Reviewed August 6, 2026 · model on record in the stance chip above.
Discussion (0). Continue with ORCID to comment.