Steven G. Johnson

> `Adjoint` is just a lazy view representing the conjugate transposed matrix This is not correct. We intentionally defined `[1,2,3]'` as a funny kind of object as a compromise between...

Unfortunately, we don't yet have an `ldiv!` implementation for the transpose of a sparse QR factorization, which is needed for this feature: #656

For reference, `Polynomials.poles` is defined here: https://github.com/JuliaMath/Polynomials.jl/blob/835a13610f1d77eff231f585d30671ed41b58fbd/src/rational-functions/common.jl#L428-L448 and does what you might expect, computing poles of a rational function. It has a docstring, but it should be added to the...

I'm not sure why it should be unexported, if Polynomials.jl supports rational functions? If you have a conflict, you can simply use it as `ControlSystems.poles` or `Polynomials.poles`, or explicitly import...

It can compute derivatives of rational functions, so you can use these to get the Taylor-series coefficient around any desired point. I'm not sure if there is a more efficient...

The main reason for the `LowerTriangular` and `UpperTriangular` types is to support special methods that only work for triangular matrices. Special matrix types for $k \ne 0$ would need to...

Should we instead fix the code to match the docstring?

This is documented in the `eigen` docstring: > Some special matrix types (e.g. `Diagonal` or `SymTridiagonal`) may implement their own sorting convention and not accept a `sortby` keyword. and >...

In general, you can't rely on BLAS implementations doing sums in anything better than the naive order. Since the norm is a BLAS1 operation, however, we arguably don't gain much...