D078 - The census now states its own limit, in the report, where a coverage figure gets quoted

Status: decided, closing BACKLOG §7.

Presence is not behaviour. A platform resolving every symbol to a stub that returns success scores full marks on the census and is useless, and the census cannot tell the difference - it reads addresses and never calls anything.

That stopped being hypothetical when the current generation's graphics interface was censused: a previous-generation emulator reports all 87 symbols present, through a generic stub, for an interface it has no implementation of.

900-surface/presence-is-not-behaviour reads the totals 007-responsive gathers - the only part of this program that knows the difference - and states the gap as the second line of the census section, ahead of the numbers it qualifies.

Its verdict is never pass. It does not test the platform, it qualifies a section, so a green line would be one more thing to add up and one more reason to stop reading. It reports partial and escalates its wording with the evidence: 22 of 54 probed functions responding gets "most probed functions are stubs; read this census as an upper bound, not a coverage figure". Zero responding is a fail.

The backlog item said the report "should probably say so more loudly than a line in the section purpose". The reason that matters: a caveat in prose is read once, and a number in the report is read every time somebody quotes a coverage figure - which is exactly the moment the caveat is needed.