ei refers to the residuals
The Huber loss function penalizes outliers and increases as a quadratic for small errors and linearly for large errors. In the JMP implementation, k = 2. For more details about robust fitting, see Huber (1973) and Huber and Ronchetti (2009).