Flexible limits for 2-categories (Q1262939): Difference between revisions

When looking at elements of a set, the primitive notion of ``sameness'' is equality. When working within a category, isomorphic objects are considered the same. However, categories themselves are considered the same when they are equivalent. Thus, the constructions within a 2- category which are of interest are those stable under equivalence. In particular, this applies to limits. Thus the appropriate limits for a 2- category are the weighted bilimits of the reviewer [``Fibrations in bicategories'', Cah. Topologie Géom. Différ. 21, 111-159 (1980; Zbl 0436.18005); Correction, ibid. 28, 53-56 (1987; Zbl 0622.18005)]. However, it quite often happens that certain limits in a 2-category can be constructed with the stricter universal property that determines them up to isomorphism, not just equivalence. When this occurs it is possible to make use of this extra precision to avoid carrying around the extra baggage of coherence. The flexible limits defined here are of the stricter kind, yet provide instances of the corresponding bilimit; a 2-category which admits small flexible limits automatically admits small bilimits. Equalizers and pullbacks are not flexible, while comma objects are. Furthermore, flexible limits have a reasonable theory, they exist in a large class of 2-categories which includes many important examples, and the third author has proved that every bicategory with finite bilimits is biequivalent to a 2-category with finite flexible limits.