Bring News from Nowhere chapters 1 through 6 into one convention

[dcellison]

The six translated chapters of News from Nowhere were compared against each other rather than against Morris again. Each chapter was internally coherent; the sequence was not. Three conventions split at the chapter 3 and 4 boundary, with no mixing inside any chapter, which is the signature of a convention change between drafting sessions rather than anything the source does. All three are now settled in one direction.

What was already consistent

The sweep confirmed more than it repaired. All 1,054 units hold the four-line shape. Recurring nouns keep one Phi form throughout: market wisola lokue, horse kalora, carriage rolutha, river luphore, forest shelira, bridge repha, garden thepalu, children lopia, street ruela, trees shiro. Structure agrees with the editorial configuration in every chapter. Source-bound treatment is clean: no Phi token outside [a-z], and no clock, coin, or unit word rendered as a numeral. Chapter 1's nested movements are its report-frame, which Morris justifies, so they stay.

Naming, ratified toward the onyms

Chapters 2 and 3 called Dick luphonu miona, waterman, while chapters 4 through 6 called him ne kulo, and five of the seven recorded onyms appeared nowhere at all. Morris names Dick from chapter 2 and twenty times in chapter 3, so the split was not the narrator learning his name.

The onyms now run through all six chapters: ne kulo, ne selomi, ne solai, ne keruko, ne luwae, ne woru, with ne phemi still the narrator's own choice, made on the page. Thirty-eight units take an onym where Morris uses a personal name. Units where Morris uses his own epithet, waterman, weaver, dustman, ferryman, sculler, keep the description; that distinction was made per unit against each citation, not by search and replace.

Canon gains the rule the decision needs: a character in a translated work cannot accept a name, so the translation is the naming community for them. A person who recurs gets one onym drawn from the relation the source gives them, used at every mention and identical across the work, with the exact source name in the adjacent line. Where a character names themselves on the page, that choice governs instead.

Quotative framing, ratified toward chapters 1 to 3

Chapters 1 through 3 frame a continuing speech once and let the following sentences run bare. Chapters 4 through 6 re-opened a full sha ... sho for each sentence of one speech, up to ten times running: at chapter 4 lines 315 to 348, Morris writes one speech opening at "O," said he, and closing seven sentences later, and the derived English said "Guide said" seven times.

Sixty-three units become bare continuation sentences. The rule applied was narrow: strip only where Morris neither opens a new quotation nor re-attributes. Where he writes "said he" or "and said", the frame is his and stays. One run was caught mid-pass because its third unit opened a new quotation, and two units were done by hand because a narrative clause sat outside the frame.

Derived-English order, ratified toward Morris

The derived English now follows Morris's own order everywhere. Where his sentence opens with the quotation, so does the English; where he writes the attribution first, as in "Said Dick:" or "Quoth Dick:", it stays first, because that is his order too. Eighty lines in chapters 3 through 6 were turned around. Chapters 1 and 2 already read this way and are untouched.

A multi-sentence speech takes its attribution after the first sentence, which is the shape chapter 2 had already established: ("Yes. I can help," said one child. The child held a calm, alert manner that amused me.)

Sixty-five lines converted mechanically and fifteen by hand: seven where the attribution is a compound verb phrase that cannot be inverted without garbling, such as "Guide smiled merrily and said", and eight where a narrative clause sits beside the speech and the Phi's own sentence order had to be preserved.

The Phi lines are deliberately not brought into Morris's order, because there they answer to grammar rather than style. The quotative frame is near-invariant across all six chapters, 123 of 149 frames reading exactly [speaker] sha ... sho [tense] [verb], and sha opens no sentence anywhere in the corpus. Phi is SOV and modifier-first, so the speaker is the subject and the speech verb lands last; the Morris inversion has no grammatical form in the language. The same holds outside dialogue: Morris trails a subordinate clause 201 times in these chapters, and where Phi uses a clause-relator, canon's modifier-first rule pulls it to the front in 57 of them. The derived English is the layer where the source's order can be honored, and it now is.

Smaller repairs found on the way

Chapter 3 line 1167 had the weaver address Dick in the derived English while the Phi contained no vocative at all; the Phi now carries it. Chapter 3 gains the great-grandfather's real name, ne keruko, which the inherited text dropped although Morris supplies it. Chapter 5 line 423 had a derived-English line with a closing quotation mark and no opening one.

Judgment calls, flagged for rejection

The English names. ne kulo was already rendering as "Guide", so the new onyms follow that pattern from their roots: ne selomi as Weaver, ne solai as Gold, ne keruko as Sturdy, ne luwae as Grace, ne woru as Keeper. Weaver is a name for a man Morris also calls the weaver, which reads as deliberate in the same way the narrator's Guest does, but it is a call worth your eye.

Chapter 3 line 347 was left unchanged. Morris's hosts call the narrator "Guest" before he adopts it, so kona phemi is the common noun there and the name arrives ten lines later. That is a distinction the source requires.

Per-file additions ledger

File	Added
chapter_01.md	Nothing. Claimed saturated: its notes already carry the driest material and end on real turns, and its one candidate addition would have repeated what the chapter intro states. Needs your confirmation or a send-back.
chapter_02.md	The onym note gains the point that a name announced by ne gains no dictionary meaning, which is why kulo still means guide elsewhere in the same chapter.
chapter_03.md	The naming note gains the observation that the hosts have been calling him phemi all morning and the name he gives is the same sound, now his. The Boffin note gains both names the passage turns on. The limits table gains a dedicated row for the people.
chapter_04.md	The market note gains an account of one frame carrying a whole speech, which is the convention this pass established.
chapter_05.md	The limits table gains a dedicated row for the weaver's onym.
chapter_06.md	The dream note gains its missing closing turn: a dream swaps one figure for another without explaining itself, and leaves the horse where it was.

Checks

validate_examples.py 0 errors, 0 warnings. validate_sentences.py 3,302 examples, 0 errors. test_sentence_validator.py passes. build_site.py succeeds; the editorial configuration's ledger-row counts were updated for the two new table rows, and the strict contract caught that omission before this ran clean. The phonetic-neighbour baseline is regenerated and committed for the first commit's attestation changes; the quote-order commit left it unchanged, since only English lines moved.

All six source reconstructions remain exact and character-for-character: 5,214, 16,061, 19,028, 6,237, 13,932 and 14,771 normalized characters, at 77, 235, 233, 116, 273 and 120 units. No unit count changed anywhere.

Voice pass on the new prose found the pet verb carry, carries, carried used across five of the six files' additions, plus "drawn from the relation Morris gives them" and "built from the relation Morris gives its bearer" a few hundred lines apart in the same chapter, and the flagged bigram "stands as". All three are corrected; the self-template was only visible by counting across files, not by reading any one of them.

Two counts in this PR were first written from memory and corrected against measurement before merge: the onym and frame-strip totals in the development log, and the eighty converted English lines. Both records carry the measured figures.

[dcellison]

The derived English now follows Morris's own order throughout, replacing the difference this PR originally left open.

Where his sentence opens with the quotation, so does the English. Where he writes the attribution first, as in "Said Dick:" or "Quoth Dick:", it stays first, because that is his order too. Eighty lines in chapters 3 through 6 were turned around; chapters 1 and 2 already read this way and are untouched.

A multi-sentence speech takes its attribution after the first sentence, which is the shape chapter 2 had already established: ("Yes. I can help," said one child. The child held a calm, alert manner that amused me.)

Sixty-five lines converted mechanically. Fifteen were done by hand: seven where the attribution is a compound verb phrase that cannot be inverted without garbling, such as "Guide smiled merrily and said", and eight where a narrative clause sits beside the speech and the Phi's own sentence order had to be preserved. Four of my own chapter 3 lines from the first commit were among them, since I had moved them the wrong way while ratifying the onyms.

Checks: validate_examples.py 0 errors and 0 warnings, validate_sentences.py 3,302 examples and 0 errors, unit tests pass, site builds. All six source reconstructions remain exact and no unit count moved. The phonetic-neighbour baseline is unchanged, since only English lines moved.

One correction to the record: I first wrote seventy-nine converted lines from memory. The measured figure is eighty, and both the development log and the handoff carry the measured one.