Chris Coey

Given how long it takes to add a cone and associated bridges, I lack the appetite for adding the CBFv3 n-dimensional power cone of questionable utility, but I will try...

This sounds like too many cones some of which aren't necessarily very practically useful in modeling. If we are to add a cone like CBF's power cone with the ||x||_2...

We have had several cones that only one unreleased solver supports for a while now, such as the geometric mean cone, the logdet and rootdet cones. They are useful even...

I've updated the issue title to narrow the scope

And here is a proposed definition: https://github.com/JuliaOpt/MathOptInterface.jl/pull/997/files#diff-0ffd5beb53539c0ff8252e848c93469eR245 An important question is whether to type the cone with ``T`` and the exponent vector with ``Vector{T}``. This is what is done for...

@blegat re: your comment https://github.com/JuliaOpt/MathOptInterface.jl/pull/997#discussion_r363411539, I asked above: > An important question is whether to type the cone with T and the exponent vector with Vector{T}. This is what is...

OK maybe I misunderstood. So can type of the exponent be different from the type of the real used in the VAF? i.e. is `VAF{T1

OK thanks, that wasn't obvious to me. In that case there is no issue. I'll amend the cone definition to be `...Cone{T}` with `exponents::Vector{T}`

It's a useful discussion, but if Benoit agrees then it can be closed

LGTM - pretty straightforward.