Malfatti's problem serves a miniature example of broader, often torturous, mathematical evolution. It has started with an apparently practical problem: to cut out of a triangular prism of a material such as marble, three circular columns of the greatest possible volume. The problem (now dubbed the marble problem [Martin, p.92]) has been published in 1803 by G. Malfatti. The problem is naturally reduced to a planar problem of packing into a given triangle three circles of the largest total area. In his solution, Malfatti assumed that the three circles in the marble problem must touch each other and each touch two sides of the triangle. Thus he reduced the problem to the one that now bears his name, Malfatti's problem [Dörrie, 147]:
It is said that a century earlier Jacob Bernoulli considered the problem for an isosceles triangle, which, if true, would have enhanced its standing in the mathematical community. (The problem also occurs in Japanese temple geometry where it is often attributed to Chokuyen Naonobu Ajima, 1732-1798.) The problem received a serious boost in 1826 when J. Steiner published (without a proof!) an elegant solution, probably with a purpose of demonstrating the power of his methods. Following is a short history of further development [Bottema]:
In 1929 Lob and Richmond showed by counterexample that in the case of an equilateral triangle the Malfatti circles (the triad of circles that solves Malfatti's problem) are not optimal in the sense that they do not answer the marble problem. Lob and Richmond showed that a different circle configuration that consists of the incircle and two circles packed in the corners of the triangle beats Malfatti's construction. In 1967, M. Goldberg gave a graphical evidence that, as a matter of fact, the Malfatti circles never solve the original marble problem. In 1992 (appeared in English in 1994), Zalgaller and Los gave a complete solution to the latter.
When the three circles in Steiner's generalization are pairwise tangent externally, the problem is trivialized by inversion [Bottema]. An inversion with center in one of the points of tangency, inverts two circles into two parallel lines: