Not signed in

Want to take part in these discussions? Sign in if you have an account, or apply for one below

• Username
• Password
• Remember me

• Sign in using OpenID
• Remember me

• Forgot your password?
• Apply for membership

2-category 2-category-theory abelian-categories adjoint algebra algebraic algebraic-geometry algebraic-topology analysis analytic-geometry arithmetic arithmetic-geometry book bundles calculus categorical categories category category-theory chern-weil-theory cohesion cohesive-homotopy-type-theory cohomology colimits combinatorics complex complex-geometry computable-mathematics computer-science constructive cosmology definitions deformation-theory descent diagrams differential differential-cohomology differential-equations differential-geometry digraphs duality elliptic-cohomology enriched fibration foundation foundations functional-analysis functor gauge-theory gebra geometric-quantization geometry graph graphs gravity grothendieck group group-theory harmonic-analysis higher higher-algebra higher-category-theory higher-differential-geometry higher-geometry higher-lie-theory higher-topos-theory homological homological-algebra homotopy homotopy-theory homotopy-type-theory index-theory integration integration-theory k-theory lie-theory limits linear linear-algebra locale localization logic mathematics measure-theory modal modal-logic model model-category-theory monad monads monoidal monoidal-category-theory morphism motives motivic-cohomology nforum nlab noncommutative noncommutative-geometry number-theory of operads operator operator-algebra order-theory pages pasting philosophy physics pro-object probability probability-theory quantization quantum quantum-field quantum-field-theory quantum-mechanics quantum-physics quantum-theory question representation representation-theory riemannian-geometry scheme schemes set set-theory sheaf simplicial space spin-geometry stable-homotopy-theory stack string string-theory superalgebra supergeometry svg symplectic-geometry synthetic-differential-geometry terminology theory topology topos topos-theory tqft type type-theory universal variational-calculus

1. • CommentRowNumber1.
   • CommentAuthorDavidRoberts
   • CommentTimeSep 30th 2010
   • (edited Oct 1st 2010)
   • PermaLink
   Author: DavidRoberts
   Format: MarkdownItexThis is to split off from the <a href="">other discussion</a> a talk about $C$-torsors for $C$ not a groupoid. The third of Moerdijk's conditions (at [[torsor with structure category]]) is stated thus: >(**freeness**) a parallel pair $u_1,u_2: c\to c'$ of morphisms in $C$, may induce coalescence $E(u_1)(\alpha)=E(u_2)(\alpha)$ for some $\alpha\in E(c)_x$ only if there is a morphism $w:b\to c$ and $\zeta\in E(b)_x$ such that $u_1\circ w = u_2\circ w$ and $E(w)(\zeta)=\alpha$. the phrase 'may induce coalescence' is a bit confusing. If it is not incorrect, I would change it to >(**freeness**) for a parallel pair $u_1,u_2: c\to c'$ of morphisms in $C$, $E(u_1)(\alpha)=E(u_2)(\alpha)$ for some $\alpha\in E(c)_x$ implies there is a morphism $w:b\to c$ and $\zeta\in E(b)_x$ such that $u_1\circ w = u_2\circ w$ and $E(w)(\zeta)=\alpha$. It seems to me that the second and third conditions are a reformulation of the conditions for a [[flat functor]].

   This is to split off from the other discussion a talk about $C$-torsors for $C$ not a groupoid.

   The third of Moerdijk’s conditions (at torsor with structure category) is stated thus:

   (freeness) a parallel pair $u_1,u_2: c\to c'$ of morphisms in $C$, may induce coalescence $E(u_1)(\alpha)=E(u_2)(\alpha)$ for some $\alpha\in E(c)_x$ only if there is a morphism $w:b\to c$ and $\zeta\in E(b)_x$ such that $u_1\circ w = u_2\circ w$ and $E(w)(\zeta)=\alpha$.

   the phrase ’may induce coalescence’ is a bit confusing. If it is not incorrect, I would change it to

   (freeness) for a parallel pair $u_1,u_2: c\to c'$ of morphisms in $C$, $E(u_1)(\alpha)=E(u_2)(\alpha)$ for some $\alpha\in E(c)_x$ implies there is a morphism $w:b\to c$ and $\zeta\in E(b)_x$ such that $u_1\circ w = u_2\circ w$ and $E(w)(\zeta)=\alpha$.

   It seems to me that the second and third conditions are a reformulation of the conditions for a flat functor.

2. • CommentRowNumber2.
   • CommentAuthorDavidRoberts
   • CommentTimeSep 30th 2010
   • (edited Oct 1st 2010)
   • PermaLink
   Author: DavidRoberts
   Format: MarkdownItexNB This is a bit of a note to myself, I don't have time to think about this now. For Street's definition (at [[torsor with structure category]]), in the diagram <img src="http://latex.codecogs.com/gif.latex? \xymatrix{ A\downarrow{a}\ar[r]^q \ar@/_2pc/[dd]_{p} \ar@%3C-.5ex%3E[d] %26 V\ar[d]^e \\ E\ar[r]_q\ar[d]_p%26 U\\ A%26 }" alt=""/>, what is the definition of the arrow $A\downarrow a \to E$? I'm guessing it exists because the span $A \leftarrow E \to U$ is a two-sided discrete fibration (of internal categories), but I haven't checked. This is a bit different to the <a href="http://www.math.mq.edu.au/~street/tcfat.pdf">2010 paper</a> (definition 4.1), but again I haven't checked.

   NB This is a bit of a note to myself, I don’t have time to think about this now.

   For Street’s definition (at torsor with structure category), in the diagram

   ,

   what is the definition of the arrow $A\downarrow a \to E$? I’m guessing it exists because the span $A \leftarrow E \to U$ is a two-sided discrete fibration (of internal categories), but I haven’t checked. This is a bit different to the 2010 paper (definition 4.1), but again I haven’t checked.

3. • CommentRowNumber3.
   • CommentAuthorzskoda
   • CommentTimeSep 30th 2010
   • PermaLink
   Author: zskoda
   Format: TextBoth things are right, the reformulation of freeness and the argument from discrete fibration.

   Both things are right, the reformulation of freeness and the argument from discrete fibration.
4. • CommentRowNumber4.
   • CommentAuthorUrs
   • CommentTimeSep 30th 2010
   • PermaLink
   Author: Urs
   Format: MarkdownItex> For Street's definition, in the diagram <img src="http://latex.codecogs.com/gif.latex? \xymatrix{ A\downarrow{a}\ar[r]^q \ar@/_2pc/[dd]_{p} \ar@%3C-.5ex%3E[d] %26 V\ar[d]^e \\ E\ar[r]_q\ar[d]_p%26 U\\ A%26 }" alt=""/>, I haven't looked at their article. Is this by any chance just the model for the lax (comma-)pullback of the point, the way we discuss at [[generalized universal bundle]], at [[Grothendieck construction]] and elsewhere?

   For Street’s definition, in the diagram ,

   I haven’t looked at their article. Is this by any chance just the model for the lax (comma-)pullback of the point, the way we discuss at generalized universal bundle, at Grothendieck construction and elsewhere?

5. • CommentRowNumber5.
   • CommentAuthorzskoda
   • CommentTimeSep 30th 2010
   • PermaLink
   Author: zskoda
   Format: MarkdownItexUrs, this diagram is from Street's article *Combinatorial aspects of descent theory* [pdf](http://www.math.mq.edu.au/~street/DescFlds.pdf), page 25 and is reproduced at [[torsor with structure groupoid]] with all the notation explained, but no further explanations. So you do not need to read the more technical paper Street-Verity to get the definition. I also wondered if it is related to the picture which you mention, but certainly not that directly. There is no universal bundle in the picture. $V\to U$ is a cover there. $A$ is an internal category. In the groupoid case, the insight is due Joyal and the following lines from Street-Verity sketch the gist of the argument ($B$ replaces $A$, $E$ replaces $C$): > André Joyal’s lectures in the Category Seminar at Macquarie University (22 October and 5 November 1980) were the other source of inspiration. André stressed the important case where $B$ is a groupoid in a fairly general category $E$ (a groupoid is a category in which every morphism is invertible; a group is a groupoid with one object). He defined a $B$-torsor to be a discrete fibration $E$ over $B$ which is locally representable and said that the relationship between $B$-torsors and cocycles could be explained in terms of the comprehensive factorization of a functor into a final functor followed by a discrete fibration.

   Urs, this diagram is from Street’s article Combinatorial aspects of descent theory pdf, page 25 and is reproduced at torsor with structure groupoid with all the notation explained, but no further explanations. So you do not need to read the more technical paper Street-Verity to get the definition. I also wondered if it is related to the picture which you mention, but certainly not that directly. There is no universal bundle in the picture. $V\to U$ is a cover there. $A$ is an internal category. In the groupoid case, the insight is due Joyal and the following lines from Street-Verity sketch the gist of the argument ($B$ replaces $A$, $E$ replaces $C$):

   André Joyal’s lectures in the Category Seminar at Macquarie University (22 October and 5 November 1980) were the other source of inspiration. André stressed the important case where $B$ is a groupoid in a fairly general category $E$ (a groupoid is a category in which every morphism is invertible; a group is a groupoid with one object). He defined a $B$-torsor to be a discrete fibration $E$ over $B$ which is locally representable and said that the relationship between $B$-torsors and cocycles could be explained in terms of the comprehensive factorization of a functor into a final functor followed by a discrete fibration.

6. • CommentRowNumber6.
   • CommentAuthorUrs
   • CommentTimeSep 30th 2010
   • (edited Sep 30th 2010)
   • PermaLink
   Author: Urs
   Format: MarkdownItexOh, from that article. Good, then it's just our notion of generalized universal bundle: So $U$ is the base space and $V \stackrel{}{\to} U$ the resolution. The anafunctor that gives the cocycle is $$ \array{ V &\stackrel{a}{\to}& A \\ \downarrow \\ U } \,. $$ We then first form the universal $A$-bundle by the pullback $$ \array{ \mathbf{E}A &\to& A_0 \\ \downarrow && \downarrow \\ A^I &\to& A } $$ and pull that back along our anafunctor to get the $A$-bundle that it classifies $$ \array{ P &\to& \mathbf{E}A \\ \downarrow && \downarrow \\ V &\stackrel{a}{\to}& A \\ \downarrow \\ U } \,. $$ Observe that here $$ P = \coprod_{x \in A} (const_x \downarrow a) $$ The remaining precomposition operation $$ Mor(A) \times_{t, s} P \to P $$ is the "principal" $A$-action on the bundle. So Street's $(Id_A \downarrow a)$ is the weak quotient $P//A$ . (That's why his "total space" object has that map to $A$.) If one wants to embed this into a good theory one cannot but demand that $V \to U$ is a weak equivalence. That means, as David R has been amplifying, that as long as the base space is just a space or an orbifold, $V$ will be groupoidal and the cocycle factor through the core of $A$ $$ a : V \to Core(A) \hookrightarrow A \,. $$ So then that total space $P$ is indeed precisely the associated bundle to the $Core(A)$-principal bundle. This is an important special case of higher bundle theory. For instance the central extensions of groupoids which people like to call nonabelian bundle gerbes are precisely such pullbacks for $A = Grpd$ and $a$ factoring through the canonical inclusion $\mathbf{B}AUT(H) \hookrightarrow Grpd$ . (On the other hand the true total space of the underlying _principal_ $AUT(H)$-2-bundle is that nonabelian bundle gerbe semidirectly multiplied with its band). Now, a truly new step into unexplored territory is obtained as soon as base space $U$ is no longer groupoidal, but taken to be a genuine category. (Such as the fundametal category of a Lorentzian spacetime.) In that case something new will happen. A notion of fiber bundle where fibers may change non-isomorphically _as time proceeds_ . I have no hint that the need for this has been observed secretly in physics anywhere, but it is a possibility.

   Oh, from that article. Good, then it’s just our notion of generalized universal bundle:

   So $U$ is the base space and $V \stackrel{}{\to} U$ the resolution. The anafunctor that gives the cocycle is

   $$\array{ V &\stackrel{a}{\to}& A \\ \downarrow \\ U } \,.$$

   We then first form the universal $A$-bundle by the pullback

   $$\array{ \mathbf{E}A &\to& A_0 \\ \downarrow && \downarrow \\ A^I &\to& A }$$

   and pull that back along our anafunctor to get the $A$-bundle that it classifies

   $$\array{ P &\to& \mathbf{E}A \\ \downarrow && \downarrow \\ V &\stackrel{a}{\to}& A \\ \downarrow \\ U } \,.$$

   Observe that here

   $$P = \coprod_{x \in A} (const_x \downarrow a)$$

   The remaining precomposition operation

   $$Mor(A) \times_{t, s} P \to P$$

   is the “principal” $A$-action on the bundle. So Street’s $(Id_A \downarrow a)$ is the weak quotient $P//A$ . (That’s why his “total space” object has that map to $A$.)

   If one wants to embed this into a good theory one cannot but demand that $V \to U$ is a weak equivalence. That means, as David R has been amplifying, that as long as the base space is just a space or an orbifold, $V$ will be groupoidal and the cocycle factor through the core of $A$

   $$a : V \to Core(A) \hookrightarrow A \,.$$

   So then that total space $P$ is indeed precisely the associated bundle to the $Core(A)$-principal bundle.

   This is an important special case of higher bundle theory. For instance the central extensions of groupoids which people like to call nonabelian bundle gerbes are precisely such pullbacks for $A = Grpd$ and $a$ factoring through the canonical inclusion $\mathbf{B}AUT(H) \hookrightarrow Grpd$ . (On the other hand the true total space of the underlying principal $AUT(H)$-2-bundle is that nonabelian bundle gerbe semidirectly multiplied with its band).

   Now, a truly new step into unexplored territory is obtained as soon as base space $U$ is no longer groupoidal, but taken to be a genuine category. (Such as the fundametal category of a Lorentzian spacetime.) In that case something new will happen. A notion of fiber bundle where fibers may change non-isomorphically as time proceeds . I have no hint that the need for this has been observed secretly in physics anywhere, but it is a possibility.

7. • CommentRowNumber7.
   • CommentAuthorTim_Porter
   • CommentTimeSep 30th 2010
   • (edited Sep 30th 2010)
   • PermaLink
   Author: Tim_Porter
   Format: MarkdownItexsee my comment on the directed spaces thread! I end up with almost the same question! I think it is linked to extensions of categories.

   see my comment on the directed spaces thread! I end up with almost the same question! I think it is linked to extensions of categories.

8. • CommentRowNumber8.
   • CommentAuthorUrs
   • CommentTimeSep 30th 2010
   • PermaLink
   Author: Urs
   Format: MarkdownItex> I think it is linked to extensions of categories. Yes, exactly! Principal $\infty$-bundles are, after all, exactly extensions of $\infty$-groupoids, namely homotopy fibers = $(\infty,1)$-fibers of maps in $\infty Grpd$ (or more generally in $(\infty,1)Sh(C)$). This is evidently the beginning of a pattern. $(\infty,1)$-bundles should be the _lax_ (or rather "comma") $(\infty,2)$-fibers of morphisms in $(\infty,1)Cat$ (or more generally in $(\infty,2)Sh(C)$). That's precisely the point, yes, bundle theory is precisely extension theory. Also known as the theory of fiber sequences. Now, the prize question is: can anyone think of a known phenomenon in theoretical physics, which would be a candidate for something that fundamentally is described by a kind of bundle on spacetime whose fibers _change in time_ ? And non-invertibly so? That would be a smoking gun indication for a manifestation of spacetime in terms of its fundamental $(\infty,1)$-category. I can't quite think of any such candidate, though. One problem of course is that fundamental physic has invertible time evolution. So I am not sure.

   I think it is linked to extensions of categories.

   Yes, exactly!

   Principal $\infty$-bundles are, after all, exactly extensions of $\infty$-groupoids, namely homotopy fibers = $(\infty,1)$-fibers of maps in $\infty Grpd$ (or more generally in $(\infty,1)Sh(C)$).

   This is evidently the beginning of a pattern. $(\infty,1)$-bundles should be the lax (or rather “comma”) $(\infty,2)$-fibers of morphisms in $(\infty,1)Cat$ (or more generally in $(\infty,2)Sh(C)$).

   That’s precisely the point, yes, bundle theory is precisely extension theory. Also known as the theory of fiber sequences.

   Now, the prize question is: can anyone think of a known phenomenon in theoretical physics, which would be a candidate for something that fundamentally is described by a kind of bundle on spacetime whose fibers change in time ? And non-invertibly so? That would be a smoking gun indication for a manifestation of spacetime in terms of its fundamental $(\infty,1)$-category. I can’t quite think of any such candidate, though. One problem of course is that fundamental physic has invertible time evolution. So I am not sure.

9. • CommentRowNumber9.
   • CommentAuthorzskoda
   • CommentTimeSep 30th 2010
   • PermaLink
   Author: zskoda
   Format: MarkdownItexThe following link related to the discussion above is temporary and the file *not* otherwise available online (djvu with Ocr) at this time * R. Street, _Characterization of bicategories of stacks_, (from LNM 962, (Category theory, Gummersbach 1981), 1982) [djvu](http://www.irb.hr/korisnici/zskoda/streetCharBicatStacksOcr.djvu)

   The following link related to the discussion above is temporary and the file not otherwise available online (djvu with Ocr) at this time

   • R. Street, Characterization of bicategories of stacks, (from LNM 962, (Category theory, Gummersbach 1981), 1982) djvu
10. • CommentRowNumber10.
    • CommentAuthorzskoda
    • CommentTimeSep 30th 2010
    • PermaLink
    Author: zskoda
    Format: MarkdownItex> One problem of course is that fundamental physic has invertible time evolution. Let's play with entropy a bit :)

    One problem of course is that fundamental physic has invertible time evolution.

    Let’s play with entropy a bit :)

11. • CommentRowNumber11.
    • CommentAuthorUrs
    • CommentTimeSep 30th 2010
    • (edited Sep 30th 2010)
    • PermaLink
    Author: Urs
    Format: MarkdownItex>> One problem of course is that fundamental physic has invertible time evolution. > Let's play with entropy a bit :) That's a good point. Maybe something like this: ordinary quantum mechanics is encoded in a Hilbert space bundle with connection (= the Hamiltonian) on the worldline. Now suppose we couple this to an environment. Then _one_ of several ways to encode the resulting decoherence is that we take partial traces that map a large Hilbert space to a smaller one $$ tr_{envir} : H_{sys} \otimes H_{envir} \to H_{sys} \,. $$ Maybe such a kind of operation is a potential candicate for a Hilbert space bundle over a directed space. But on the other hand that tracing is a somewhat discrete operation. Ah, here is another idea (will post this in the next comment...)

    One problem of course is that fundamental physic has invertible time evolution.

    Let’s play with entropy a bit :)

    That’s a good point. Maybe something like this: ordinary quantum mechanics is encoded in a Hilbert space bundle with connection (= the Hamiltonian) on the worldline.

    Now suppose we couple this to an environment. Then one of several ways to encode the resulting decoherence is that we take partial traces that map a large Hilbert space to a smaller one

    $$tr_{envir} : H_{sys} \otimes H_{envir} \to H_{sys} \,.$$

    Maybe such a kind of operation is a potential candicate for a Hilbert space bundle over a directed space.

    But on the other hand that tracing is a somewhat discrete operation.

    Ah, here is another idea (will post this in the next comment…)

12. • CommentRowNumber12.
    • CommentAuthorUrs
    • CommentTimeSep 30th 2010
    • PermaLink
    Author: Urs
    Format: MarkdownItexHere is a better idea: maybe we shouldn't be thinking about bundles whose fibers change non-invertibly. Rather we should be thinking of bundles with _connection_ whose parallel transport is no longer with values in isomorphisms. And _that_ of course has immediate physics applications: The operation $(t \to t') \mapsto (H_t \stackrel{\exp(i (t'-t) K)}{\to} H_{t'})$ is a parallel transport/ on the undirected worldline: unitary time evolution. Now switch on dissipation and replace the worldline's fundamental $groupoid$ with its fundamental category: then we may have also dissipative contributions $(t \to t') \mapsto (H_t \stackrel{\exp((t'-t) (i K - Q))}{\to} H_{t'})$ (for $K$ and $Q$ hermitean operators, hence with real spectrum). So the fibers $H_t$ are isomorphic for all $t$, but the parallel transport is no longer an isomorphism between them.

    Here is a better idea: maybe we shouldn’t be thinking about bundles whose fibers change non-invertibly.

    Rather we should be thinking of bundles with connection whose parallel transport is no longer with values in isomorphisms.

    And that of course has immediate physics applications:

    The operation $(t \to t') \mapsto (H_t \stackrel{\exp(i (t'-t) K)}{\to} H_{t'})$ is a parallel transport/ on the undirected worldline: unitary time evolution.

    Now switch on dissipation and replace the worldline’s fundamental $groupoid$ with its fundamental category: then we may have also dissipative contributions

    $(t \to t') \mapsto (H_t \stackrel{\exp((t'-t) (i K - Q))}{\to} H_{t'})$

    (for $K$ and $Q$ hermitean operators, hence with real spectrum).

    So the fibers $H_t$ are isomorphic for all $t$, but the parallel transport is no longer an isomorphism between them.

13. • CommentRowNumber13.
    • CommentAuthorzskoda
    • CommentTimeSep 30th 2010
    • (edited Sep 30th 2010)
    • PermaLink
    Author: zskoda
    Format: MarkdownItexWow, you are brave going into an unknown land...(the idea with the partial trace toward such bundles).

    Wow, you are brave going into an unknown land…(the idea with the partial trace toward such bundles).

14. • CommentRowNumber14.
    • CommentAuthorzskoda
    • CommentTimeSep 30th 2010
    • PermaLink
    Author: zskoda
    Format: MarkdownItexSo you still call that case a connection ? (I am not complaining)

    So you still call that case a connection ? (I am not complaining)

15. • CommentRowNumber15.
    • CommentAuthorUrs
    • CommentTimeSep 30th 2010
    • PermaLink
    Author: Urs
    Format: MarkdownItex> So you still call that case a connection Well, I didn't up to this point, but it would make sense to call this a connection with parallel transport along directed paths, yes. I did think about connections and higher connections on directed spaces before. In my article on AQFT I show -- slightly paraphrased in the terms we started using above -- that every parallel 2-transport on the fundamental $(2,1)$-category of a causal 2d Lorentzian manifold induced a local net of observables. For that statement it is _required_ that the 2-transport takes values in invertible 2-morphisms. The directedness of the underlying Minkowski space affects not the notion of parallel transport, but is the extra structure that allows to extract a local net of observables from it. But one could also consider parallel 2-transport on the fundamental $(2,1)$-category of a Minkowski space with values in non-invertible 2-morphisms. That would _not_ give rise to a local net of observables -- but that is good, because it would describe a "Euclidean" QFT and these do not come with local nets of observables (not in that sense of local, anyway). So, yeah. I haven't thought much about non-invertible parallel transport, but it certainly sounds like a sensible thing to consider in this context.

    So you still call that case a connection

    Well, I didn’t up to this point, but it would make sense to call this a connection with parallel transport along directed paths, yes.

    I did think about connections and higher connections on directed spaces before. In my article on AQFT I show – slightly paraphrased in the terms we started using above – that every parallel 2-transport on the fundamental $(2,1)$-category of a causal 2d Lorentzian manifold induced a local net of observables.

    For that statement it is required that the 2-transport takes values in invertible 2-morphisms. The directedness of the underlying Minkowski space affects not the notion of parallel transport, but is the extra structure that allows to extract a local net of observables from it.

    But one could also consider parallel 2-transport on the fundamental $(2,1)$-category of a Minkowski space with values in non-invertible 2-morphisms. That would not give rise to a local net of observables – but that is good, because it would describe a “Euclidean” QFT and these do not come with local nets of observables (not in that sense of local, anyway).

    So, yeah. I haven’t thought much about non-invertible parallel transport, but it certainly sounds like a sensible thing to consider in this context.

16. • CommentRowNumber16.
    • CommentAuthorDavidRoberts
    • CommentTimeOct 1st 2010
    • (edited Oct 1st 2010)
    • PermaLink
    Author: DavidRoberts
    Format: MarkdownItexI have changed the 'freeness' condition at [[torsor with structure category]] as I suggested in my #1 above. @Urs, I know $A\downarrow a$ comes with a map to $A$, the question was how it comes with a map to $E$. Let me just write it out for my own explanation, I'll put this into the entry when I'm happy with it. Let me reproduce the diagram <img src="http://latex.codecogs.com/gif.latex? \xymatrix{ A\downarrow{a}\ar[r]^q \ar@/_2pc/[dd]_{p} \ar@%3C-.5ex%3E[d] %26 V\ar[d]^e \\ E\ar[r]_q\ar[d]_p%26 U\\ A%26 }" alt=""/> So $V\to U$ here is a map in the ambient category $S$ and $A$ is a category in $S$. $disc(U)\leftarrow E\to A$ is anafunctor-like, but I'm not sure how close it is to the 'usual' notion. The 'locally trivial condition' asks that for the cover $V\to U$ (for Street a regular epimorphism, hence an element of a subcanonical singleton pretopology), there is a map $disc(V)\to A$, i.e. a map $V\to A_0$ such that $E\times_{disc(U)} disc(V) \simeq A^I \times_A \disc(V)$ ($A^I$ = category of isomorphisms) - this much is clear to me, because $A \leftarrow A^I \to A$ is the (weak!) identity 1-arrow in the 2-category of saturated anafunctors, and so represents the trivial $A$-torsor. However, it is not immediate that $A^I \times_A \disc(V)$ comes with an arrow to $E$ so as to make it the pullback. Zoran tells me my guess is true that the lifting properties inherent in $E$ being discretely fibred over $U$ and $A$ gives this map (and it had better be unique), so I think I'll have to sit down and convince myself how this works. Then there is the problem of showing how the old definition and the new one (from the 2010 paper) line up.

    I have changed the ’freeness’ condition at torsor with structure category as I suggested in my #1 above.

    @Urs,

    I know $A\downarrow a$ comes with a map to $A$, the question was how it comes with a map to $E$.

    Let me just write it out for my own explanation, I’ll put this into the entry when I’m happy with it. Let me reproduce the diagram

    

    So $V\to U$ here is a map in the ambient category $S$ and $A$ is a category in $S$. $disc(U)\leftarrow E\to A$ is anafunctor-like, but I’m not sure how close it is to the ’usual’ notion. The ’locally trivial condition’ asks that for the cover $V\to U$ (for Street a regular epimorphism, hence an element of a subcanonical singleton pretopology), there is a map $disc(V)\to A$, i.e. a map $V\to A_0$ such that $E\times_{disc(U)} disc(V) \simeq A^I \times_A \disc(V)$ ($A^I$ = category of isomorphisms) - this much is clear to me, because $A \leftarrow A^I \to A$ is the (weak!) identity 1-arrow in the 2-category of saturated anafunctors, and so represents the trivial $A$-torsor. However, it is not immediate that $A^I \times_A \disc(V)$ comes with an arrow to $E$ so as to make it the pullback. Zoran tells me my guess is true that the lifting properties inherent in $E$ being discretely fibred over $U$ and $A$ gives this map (and it had better be unique), so I think I’ll have to sit down and convince myself how this works.

    Then there is the problem of showing how the old definition and the new one (from the 2010 paper) line up.