Tim Holy

So the easier thing is to boot out the optimization and get things at the time you need them (in the right world age).

> I think [@aviatesk](https://github.com/aviatesk) has said that this whole optimization of resolving values early is invalid anyway though.. It's also been deactivated (which is partly why so much is broken)...

I wonder whether `solved=true` and "found approximate minimum least-squares solution" is really the right status message in that case?

Looking good! What I meant was the following: ```julia function reencode1!(out, types::Vector{DataType}, vals) for (T, v) in zip(types, vals) push!(out, T(v)) end return out end function reencode2!(out, typecodes::Vector{Int}, vals) for...

Just a couple of links in case they are useful: - explanation of fields: https://mmcif.wwpdb.org/dictionaries/mmcif_pdbx_v50.dic/Categories/struct_ref_seq.html - UniProt's structure page for this P00974 entry: https://www.uniprot.org/uniprotkb/P00974/entry#structure (1bpi is listed as spanning the...

Another option might be to store extra fields in the `MolecularStructure`, but the Dict is empty unless the user requests certain fields: ```julia struc = read("AAAA.cif", MMCIFFormat; headerfields = ["_struct_ref_seq"])...

Oh, nvm, I see one uses `λ` and the other `λ^2`.

Actually, I wonder if this is basically a violation-of-intent for the [generic interface](https://jso.dev/Krylov.jl/stable/generic_interface/): having the same kwarg supported by multiple solvers, but having it mean something different in different solvers.

Inspired in part by seeing the activity in #45

I agree this is a tricky issue. Fundamentally I'm not sure. Plain type-changes often aren't, but mutable to immutable is a behavior change. If we decide it's breaking, then we...