Francesc Verdugo

At least, the ghost layer has to be implemented taking into account periodicity. And perhaps other devs.

Hi @oriolcg @amartinhuertas, I would use indexing/iteration just like for the FEFunction: ```julia V = MultiFieldFESpace([V1,V2,V3]) V1,V2,V3 = V V2 = V[2] ```

> I think field_fe_space is a more intention revealing name than spaces. I would use the former, may be with a trailing s, i.e., field_fe_spaces for both packages. What do...

> If this works I'll do that and change the name in both libraries to field_fe_spaces, and we don't need the getter. If it is not needed I would not...

> The indexing/iteration functions are already implemented That's great! I didn't remember! If U[2] is available, I think it makes sense to also have length(U).

> We could create a solver from DifferentialEquation at every time step, from t_n to t_n+dt. I cannot see how to use the methods in the library as iterators, so...

> Is it possible to do what I want in DifferentialEquations.jl? If Yes, What am I doing wrong? I could not find an example in the documentation covering this (probably...

cc @ahojukka5 @KristofferC

Hi @fredrikekre IMHO, understanding how artifacts work and setting up an override looks way more complicated than reading the documentation of a function, calling the function, and restarting Julia (proposed...

> I think the Overrides.toml is pretty much deprecated. However, using preferences I think jlls already have support for this, see e.g. https://docs.binarybuilder.org/stable/jll/#Overriding-specific-products. This looks promising. Do you have a...