Do you think it is OK to merely suggest nulls, but allow any number? (Preferably not even "SHOULD be null", but "MAY be null".)

I was a somewhat strong proponent for standardizing on reduced coordinates. Hence, I will certainly provide an omdb_reduced_site_positions alongside the standard cartesian_site_positions. In reduced coordinates, it makes sense to have lattice vectors also for the non-periodic coordinates, they act as multipliers for the coordinates in the omdb_reduced_site_positions list, so, for me, those values are going to be meaningful.

Non-null values indeed make sense having considered the reduced coordinates. Is there an issue open for their inclusion? Maybe we could get reduced_site_positions added before v1.0 release?

No issue as far as I know. I've only planned it as a database-specific property. Unless multiple databases chime in an say it is very useful for them, we can standardize it next meeting.

I'd suggest either to:

1. Standardize reduced_site_positions for v1.0 and require non-null values when reduced_site_positions is present;
2. Leave reduced_site_positions for v1.1 and discourage non-null values for nonperiodic vectors until then.

What do you think?

I say, 2, but not so much discourage non-null values as saying that values MAY be null.

But if someone wants to do 1, I'll be happy to merge that PR.

But if someone wants to do 1, I'll be happy to merge that PR.

I have opened PR #206.