isAbsentValue decides on literal strings rather than the yaml null and empty class so chasing spellings cannot close it

isAbsentValue decides on literal strings rather than the yaml null and empty class so chasing spellings cannot close it

Found by the itd-189 delta security review, and separated from iss-2608301808193750 on purpose: one is a false sentence to correct now, this is the design question underneath it.

isAbsentValue compares against literal strings -- v == "!!null" and friends -- and isEmptyFlowCollection anchors on the first and last byte. The reviewer's sentence is the one to keep: the new constant is a spelling test, not a null test. !!null null and !<tag:yaml.org,2002:null> are the SAME YAML node as !!null, and the predicate accepts two of the three.

So the eleven spellings are not eleven bugs. They are one wrong altitude, and adding an eleventh case would leave a twelfth.

This is the same shape adr-56 just ruled on for the exclusion floor, one workstream over: a control recognising a construct by its spelling rather than by what it IS, closing spellings one at a time across nine rounds while the next unenumerated one stays open. The records there say it outright -- a spelling arms race needs a design, not more patterns -- and the ruling was to change the altitude rather than extend the list.

The complication here is that there is no YAML library in the module (the reviewer established this while refuting a different finding), so "decide on the tag and anchor class" is real work rather than a swapped call. That is why this is captured rather than fixed in the round that found it.

Two options for whoever takes it, neither chosen here: teach the shared scalar reader the tag, anchor and alias productions so emptiness is decided once for every field; or narrow what the gate claims to what a string comparison can honestly establish, and say so where the operator reads it. The first is adr-56's branch-2 analogue applied to this surface; the second is its branch-3.

Grounds

• pursued: we expect deciding on the node's class rather than on its spelling to close the spellings nobody enumerated at the same time as the ones that were, so no future round has to add an eleventh literal; a YAML null, empty collection or empty string that the shared reader still calls populated, or a populated value it calls empty, would show it wrong.