Математики любят говорить о том, как правильно надо преподавать математику. Мне не раз попадались высказывания в духе того, что линейная алгебра и абстрактная алгебра преподаются студентам-математикам в наше время так, что получаются выхолощенными от своей сути наборами технических определений. Британский математик Майлз Рид пишет об этом так выразительно, что захотелось процитировать (прошу прощения за длинную цитату по-английски):
"...The problem is that the abstract point of view in teaching leads to isolation from the motivations and applications of the subject. For example, differential operators are typical examples of linear maps, used all over pure and applied math, but it is a safe bet that the linear algebra lecturer will not mention them: after all, logically speaking, differentiation is more complicated than an axiom about T(av + bu), and working with infinite dimensional vector spaces would clearly needlessly disconcert the students. In the same way, if the applied people want students to study coordinate geometry in R^3, let them set up their own course - the student who understands that the applied lecturer's R^3 is an example of the algebra lecturer's vector spaces will be at an unexpected advantage. Similar examples occur at every point of contact between algebra and other subjects; under the system of abstract axioms, the algebraist is never going to take responsibility for relating to the applications of his subject outside algebra.
No subject has suffered as badly from the insistence on the abstract treatment as group theory. When I was a first year undergraduate in Cambridge in 1966, it had been more or less settled, presumably after some debate, that the Sylow theorems for finite groups were too hard for Algebra IA; since then, the notion of quotient group, and subsequently the definitions of conjugacy and normal subgroup have been squeezed out as too difficult for the first year. Thus our algebraists have cut out most of the course, but stick to the dogma that a group is a set with a binary operation satisfying various axioms. Groups can be taught as symmetry groups (geometric transformation groups), and the abstract definition of group held back until the student knows enough examples and methods of calculation to motivate all the definitions, and to see the point of isomorphism of groups.
The schizophrenia between abstract groups and transformation groups comes to the surface in some amusing quirks - for example, the textbooks that define an "abstract group of operators", or the students (year after year) who insist that the binary operation GxG->G on a group should satisfy closure under (g1,g2) -> g1g2 as one of the group axioms. It seems to me that the abstract approach has weaknesses even within the framework of abstract algebra. In recent years, the Warwick 3rd year has featured a course on Lie algebras. I've no doubt that the course is extremely well given, but it's still possible to find students who get good grades, and know all the bookwork in the course, but who still don't know that nxn matrices over R with bracket [A, B] = AB - BA is an example of a Lie algebra, and R^n with Av = matrix times a vector an example of a representation or a module. The student who knows just this one example can make good sense of the entire course. Of course, given a chance, any self-respecting geometer, applied mathematician or physicist would insist on muddling things up by differentiating the group law at the origin, and explaining what happens to the associative law, etc. Is it conceivable that there are people about who introduce the Jacobi identity as a bald axiom?"