“…The localization principle put forward in [18] says that for all "natural orthogonal systems" the kernels {L n (x, y)} decay at rates faster than any inverse polynomial rate away from the main diagonal y = x in E × E with respect to the distance in E. This principle is very well-known in the case of the trigonometric system (and the Fourier transform) and not so long ago was established for spherical harmonics [15,16], Jacobi polynomials [1,17], orthogonal polynomials on the ball [18], and Hermite and Laguerre functions [2,4,19,9]. Surprisingly, however, the localization principle as formulated above fails to be true for tensor product Jacobi polynomials and, in particular, for tensor product Legendre or Chebyshev polynomials, as will be shown in §10.…”