Why is the definition of flat $(\infty, 1)$-functor in nlab (=left exact in HTT, meaning that the pullback along $f: A\to B$ preserves left fibration with cofiltered total space) correct? I'm interested in the case $B=\Spaces$, and I believe that the general case is defined representably using this. So I mean why is it better than the competing weaker (HTT 5.3.2.11) property that the 'category of elements' = unstraightening
of the functor $f: A \to \Spaces$ is cofiltered? Which one is equivalent to being a filtered colimit of representable functors, or its Yoneda extension being left exact?
of the functor $f: A \to \Spaces$ is cofiltered? Which one is equivalent to being a filtered colimit of representable functors, or its Yoneda extension being left exact?