Should Smith y V, John keep the V in the family name, as Smith Jones y V, John does? #1093

Before a family comma, a single-letter connective standing right in front of a one-letter roman numeral counts as a word to spare for that numeral, so the numeral reads as the generation. One more word in front and the opposite happens: the connective joins the numeral into the family name.

The comma-less twins read the suffix in every release: Smith y V reads given Smith, family y, suffix V, and Smith Jones y V given Smith, middle Jones, family y, suffix V.

A two-letter numeral shows the same split on master, which unlike V has not shipped yet: Smith y VI, John reads suffix VI since b6e2b97 (#893; 2.8.2 reads family Smith y VI), while Smith Jones y VI, John reads family Smith Jones y VI.

Where each reading comes from

Two counts read the same part, and they read the numeral in opposite ways.

1. group, P3's carve-out (connectives). P3 keeps a single-letter connective unjoined in a name of fewer than four root words. In a family part the given part's words count too (additional), and P3's numeral clause leaves out a numeral that the trailing fork reads as the generation ("no word of the three"):
   • Smith y V, John: Smith, y, V and John give 4, less V gives 3. That is below 4, so y stays a piece of its own (letter_stays=True).
   • Smith Jones y V, John: 5 − 1 = 4, so y joins. Its right-hand neighbour is the V the count had just left out as the generation, so the join takes it: [Smith] [Jones y V].
2. assign, the family part's numeral site (rules/C1's numeral clause, _family_part_numeral, added by Ask one predicate for every roman-numeral rule (#752, phase 2f) #886 for Should John Smith V, M.J. keep V as the suffix, as John Smith Jr., M.J. keeps Jr.? #774). It asks S10's words-to-spare count (the trailing roman numeral) over the pieces in front of the numeral, and the unjoined y is one of those pieces. Smith y V has two pieces in front of V, so V reads as the suffix. In Smith Jones y V the numeral is inside the join, so the site never sees it.

Bisected with the released wheels and then by commit: Smith y V, John moved in 1f0b07c (#886, phase 2f of #752), the commit #1080 bisected to. Before it, the family walk kept V in the family (the initial veto), and the join decision is the same on both sides of the commit. The VI row moved in b6e2b97 (#893), when P3's count began leaving VI out.

Is it already accepted?

Not for this shape.

• decisions/C1's Should John Smith V, M.J. keep V as the suffix, as John Smith Jr., M.J. keeps Jr.? #774 entry accepts that "a part holding a connective surname reads its one-letter numeral as its comma-less twin does only where the two join the connective alike". Its row is Vega y Lopez V, Mary, where the join stands in front of the numeral. It says nothing of a connective right in front of the numeral, unjoined in one case and joined across the numeral in the other.
• rules/C3's connective clause (a connective join is not one word to the comma's count) is about the comma decision's count, not the numeral's.
• Count a family part's particles out of its numeral count (#1080) #1091 (fixing Why does "SMITH VD V, JOHN" lose the V from the family name when "SMITH V, JOHN" keeps it? #1080) makes the same site count a piece made wholly of particles as no word, because "the comma has named the part a surname". It leaves the connective out on purpose, as a separate rule.

No local corpus holds the shape: a connective followed by a numeral before a family comma appears in no corpus.jsonl among the cached local corpora (fec-individuals, ctgov-officials, tse-donors, psc-individuals, ukec-individuals, dof-signatures and the rest, grepped 2026-11-05). The only match is Dr Padmavathi I V, MBBS,MD, which has a suffix comma.

Proposal (in rule shape)

When: a single-letter connective (P3) stands right in front of a one-letter roman numeral that ends the family part before a family comma, as in "Smith y V, John".

Then: the connective is no word to spare for the numeral (S10's count). The numeral is the name the connective links, and the part reads family whole.

"Smith y V, John"           →  family="Smith y V"
"Silva e V, João"           →  family="Silva e V"
"Smith Jones y V, John"     →  family="Smith Jones y V"
"Smith y V, John"           →  ambiguities=()

Because: the comma has named the part a surname, and a connective in a surname links two names. Read as the generation, the V leaves the y linking nothing, the family Smith y, which no writer means. P3 already says this of a connective that is also generational vocabulary: "a connective with nothing to its right is connecting nothing". It is the reasoning #1080 used for a particle, and it gives back the 1.4.0 – 2.6.0 reading.

Except:

• A letter written with a period stays a name word (T5), as it does today.

  "Smith y V., John"          →  family="Smith y V."
  

• A name written wholly in one case already keeps the letter, the Y in front being written as an initial (T4).

  "SMITH Y V, JOHN"           →  family="SMITH Y V"
  

• A lower-case numeral in a name of mixed case stays the generation (S10's written-case clause).

  "Smith y v, John"           →  suffix="v"
  

Accepted: the family part then departs from its comma-less twin (Smith y V keeps suffix V), as #1091's John van der V, M.J. does for a particle. The comma-less name has a given name for the connective to follow; a family part has none.

Open questions

1. Direction. The alternative is the comma-less twin: both rows read suffix V, and group's join stops taking as its right-hand word a numeral its own count has just left out as the generation (Smith Jones y V, John → family Smith Jones y, suffix V). That keeps rules/C1's "read as the comma-less name reads it" intact, at the cost of a family name ending in a bare connective. The proposal above takes the other side. Whichever is chosen, the disagreement between the count and the join in Smith Jones y V, John is one question answered two ways, and that part needs fixing either way.
2. Two or more letters. VI asks no words-to-spare count, so the proposal does not reach Smith y VI, John (suffix VI on master, family whole in 2.8.2). Should a connective in front pull a shape-read numeral into the family too? If so, it reverses Should a numeral before the comma count as a name word? (Dumas VI, Jr. gives no family name, Dumas III, Jr. gives family Dumas) #893's effect on this one shape. If not, Smith y VI, John and Smith Jones y VI, John keep disagreeing.
3. Where it is stated. S10's family-comma Precedence bullet, beside Count a family part's particles out of its numeral count (#1080) #1091's particle clause, or a P3 clause, since it is the connective's job that decides it?

Out of scope

• Smith i V, John (suffix i V) is P3's clause for a connective that is also generational vocabulary. It moved in the same commit, and 1.4.0 – 2.6.0 read family Smith V, suffix i, which is no better.
• Multi-letter connectives (and, &) join at any length and are not affected: Smith and V, John reads family Smith and V.