Comment by snicker7

Despite what the article claims, normality is not actually an assumption of linear regression. It is "required" for doing F-tests (the F-distribution being related to the normal distribution), but it is not required for proving that the regression coefficients are consistent.