> IMO we should be dropping the glance image constraint. It is a massive overreach.
Do you mean inside the resource schema? If so then I tend to agree - we should allow template authors to express those constraints on their input parameters, but having dynamic validation paths like this inside statically validated resource schema interfaces does seem like a step too far.
The patch I posted solves the issue for the specific case of count=0 for ResourceGroup, but wider discussion is probably needed before we remove all resource schema custom contraints.
> IMO we should be dropping the glance image constraint. It is a massive overreach.
Do you mean inside the resource schema? If so then I tend to agree - we should allow template authors to express those constraints on their input parameters, but having dynamic validation paths like this inside statically validated resource schema interfaces does seem like a step too far.
The patch I posted solves the issue for the specific case of count=0 for ResourceGroup, but wider discussion is probably needed before we remove all resource schema custom contraints.