Isolation

Cases that run in sequence can contaminate each other: a case that creates a booking changes what the next case sees. Isolation says whether RigorRun could stop that.

	What it means
RESET	The cases were observed to start from the same state. RigorRun performed the reset itself and compared.
PARTIAL	Reset ran, and only some of the state came back the same.
DECLARED	A reset was nominated, and nothing has run it twice and compared the results. Believed, not observed.
NONE	There is no way to put the system back.

DECLARED is not a weaker RESET

This overclaimed until 0.2.0

RigorRun reported RESET whenever any reset was configured, though nothing had ever run the reset twice and compared the results. It was reporting an observation it had never made.

It now says RESET only where the reset is RigorRun’s own machinery — an in-process snapshot, or a container thrown away — and DECLARED where it is a tool the operator nominated. Most real connected systems are therefore DECLARED.

With no reset at all

A system with NONE still works, and produces a smaller suite:

• Repeated mutating cases are switched off rather than quietly run.
• Every result says the cases could have affected each other.
• Suite quality checking is unavailable, because it runs the suite several times and most of those runs write.

That is a fair trade, and it is better than a suite that looks complete and is not.