Shyue Ping Ong

I will weigh in a little. My recommendation is: i. No need for MPForceRelax. It is pretty dumb to have a MPForceRelax that replicates MPRelax but with just the additional...

I don't think (iv) will change probability of convergence much. This is after in the static calculation, where you are already starting from a well-relaxed geometry as well as CHGCAR....

Again, when I wrote *comments*, I meant useful ones. Like saying `we use EDIFF instead of EDIFF_PER_ATOM for , as specified in MPStaticSet (ref link)`. No one asked to write...

Input set can be specific as a FULL string. We regularly write these as mylab.sets.MyOwnSet. It is already the way it is implemented in the MontyDecoder. These will allow people...

OK. Start by moving surface first then.

I am not sure but the information might be in the transformations history. @mkhorton can confirm

@mkhorton @shyamd Seriously the mapidoc is very out of date. For example, `anonymous_formula` is still there even though it has been deprecated in favor of `formula_anonymous`. The key `icsd_ids` does...

I would say that there are three levels. Tier 1 is as defined by Anubhav. Tier 2 should be “fake” VASP. The idea here is to ensure that the logic...

I would exclude Tier 3 from CI testing.

@mkhorton I disagree that "fake" VASP adds engineering burden. I would say that a basic integration test would be running a full Firework, but with "fake" output files being put...