Ronny Bergmann

> Note that YPm, YPn, gps, and hps would be stored in AugmentedLagrangianGrad to avoid unnecessary allocations, and ConstrainedManifoldObjective would have to store the array representation type. But then you...

Ah and most the hustle I went through in the last rewrite (and all thinking last week) was to avoid having to regenerate the Power manifold on every function call,...

Hm, what would still allocate a power manifold in every call? Sure sorting just the power representation is maybe ok, but I would prefer (similar to the manifold not being...

I thought about this for a while, and it seems most-unified to put the `p=` as a keyword argument instead. That way the user can provide different parts of the...

The mutating keyword is the one thing I to not like so much, but I also do not like to move mandatory things like the Hessian to keywords. So I...

I think I have an idea that even makes the objective a bit more flexible and maybe even reduces code duplication. Similar to the `VectorialFunction`, where now every constraint has...

I am not yet so sure how that helps, but I can think about how that might be meant.

But you might be right that function types are maybe overcomplicating things – the main idea comes from maybe indeed “typing the Hessian” (though not in this but the other...

I do not yet see how your dispatch solves the problem with the ambiguity to `difference_of_convex_algorithm(M, mdco, p; kwargs...)` In most cases the user would not need the `HessianFunction` type...

So there are two problems: sometimes objectives & start point more often Hessian and start point. But I can check and write them out here (or in the initial post)...