Meaning of “a mapping preserves structures/properties”

1. Sometimes I see something like "a mapping preserves the structures of its domain and of its codomain". From Wiki about morphisms in category theory:

a morphism is an abstraction derived from structure-preserving mappings between two mathematical structures. The notion of morphism recurs in much of contemporary mathematics. In set theory, morphisms are functions; in linear algebra, linear transformations; in group theory, group homomorphisms; in topology, continuous functions, and so on.

I was wondering why the structure-preserving mappings between two topological/measurable spaces are defined by the "inverse" of the mapping, while the structure-preserving mappings between two groups/vector spaces are not?

Why are the structure-preserving mappings between two topological spaces chosen to be continuous mappings instead of open mappings?

2. I also see that "a mapping preserves some property of subsets, points or whatever". Such as

Continuous linear mappings between topological vector spaces preserve boundedness.

According to Brian's reply to my earlier question, this quote should be understood as "under a continuous linear mapping, the image of any bounded domain subset is also a bounded codomain subset", not as "under a continuous linear mapping, the inverse image of any bounded codomain subset is also a bounded domain subset".

I wonder why? It seems at first to me like how continuous mappings preserve topologies, but it is actually in the same way as how group homomorphisms preserve group structures.

Thanks and regards!

• Note that for number two, if the linear map is a bijection then both statements are true. Furthermore, bijective continuous linear maps between topological vector spaces are isomorphisms, which is what is typically meant by a map that "preserves structure". – Alex Becker Feb 25 '12 at 3:49
• It's worth noting that while maps are often defined going backwards or forewards, other information often travels the opposite way. The continuous image of a compact set is compact, and the preimage of a normal subgroup is normal, for example. – Brett Frankel Feb 25 '12 at 3:51
• @AlexBecker: My understanding is that the bijective case is for isomoprhism, but is not necessary for "preserving structures/properties". For example, homomorphisms and isomorphisms between two groups. – Tim Feb 25 '12 at 3:51
• In order to view topological maps (i.e., continuous maps) as "preserving a structure", you really need to think of them in terms off preserving the notion of "closeness", not the notion of "open sets". It just so happens that the right way to say "$f$ sends points that are close-to-one-another to points that are close-to-one-another" is via inverse images when you consider open sets. To define it in terms of direct images, you consider instead the filter of neighborhoods of a point. – Arturo Magidin Feb 25 '12 at 3:52
• @Tim: You're right about Lebesgue functions pulling Borel sigma-algebras back to Lebesgue sigma-algebras, but the analogy is not quite right here, since in category theory we want to be able to compose arrows. Compositions of Borel-measurable functions are Borel-measurable, but compositions of Lebesgue-measurable functions are generally not Lebesgue-measurable. – Brett Frankel Feb 25 '12 at 4:26

Top is the category of topological spaces and continuous maps simply by definition; topology typically deals with continuous maps, making this category the most important one, and thus by convention it's the one meant when saying "the category of topological spaces".

(aside: other conventions on what Top or "the category of topological spaces" stands for are far more likely to disagree on what the objects are, rather than the morphisms. e.g. to make the objects be merely the compactly generated Hausdorff spaces)

You can, of course, make all sorts of other categories. The category of topological spaces and open maps is a perfectly reasonable category to make; it's just less useful.

It takes a bit to get used to, but category theory rejects the mindset that mathematics is about objects, with the mappings between objects being a derived notion. Instead, you need to consider objects and mappings as equals -- or even to consider the objects superfluous.

On that last point, my favorite example of a category whose emphasis is on the morphisms is matrix algebra. The set of all matrices, with composition defined by multiplication, form a category. (with addition, you get an Abelian category) The objects of this category really play no role beyond bookkeeping to say which matrix products are defined.

(This category is, of course, equivalent to the category of fintie-dimensional vector spaces and linear maps)

• +1 Thanks! (1) Why in the category of matrices, the objects "matrices" are superfluous? (2) How about measurable mappings preserving sigma algebras? – Tim Feb 25 '12 at 4:28
• In the category formed by matrix algebra, the matrices are the morphisms, not the objects. If you insist on there being objects, in the simplest version of this category, the objects would be numbers. So an $m \times n$ matrix is a morphism from $n$ to $m$ (or vice versa, depending on your conventions). – Hurkyl Feb 25 '12 at 9:29

The question actually boils down to:
"Why do we define continuous maps as the way we do?!?"