Repository path: framework/v0.3/entries/HB-meter-and-line.md · rendered 2026-09-09

Where the source is in a fixed measure and rhymes at every line end, and the translator must sign or refuse a formal contract before the first line

Standing. Written at the project's close (2026-09-09, S257) as a consolidation of the record without the translation limb the procedure's step 6 requires — no fresh passage was translated under this entry, so it stays status: draft and its §5 Application reads 'not applied'. The evidence is X1a/X1b/X2 only: two published hands of one author read whole (Leaf 1898 audited at 28 odes; Payne 1901 counted for syllables and sense placement, his metre unaudited — his OCR is unusable) and five lead poems (seven arms) under four regimes with frozen logs; every craft observation and every gold key is the lead's own coding, internal-judgment-only. Tier D is NOT PASSED and is EXHAUSTED (config/models.md, RS-20260906-tierD-verdict-v3): nothing here says one rendering is better than another, and no sentence says a reader hears a metre or a rhyme. Consolidated from v0.2 §7.43, §7.47, §7.48, §7.51 and the line-end half of §7.49, against RS-20260908-chime-slot's corrected number; §7.49's by-radif robustness is written as unresolved.

1. The problem

A Hafez ghazal is in one quantitative measure from first line to last (24 measures), monorhymed at both hemistichs of the first bayt and the second of every later one, and in 305 of 495 ghazals the rhyme is followed by a radif — the same word standing last at every rhyming position (family C). Each line end is spoken for up to three times: by the measure's final feet, by the qāfiya, by the radif. English is stress-timed, renders quantity only by stress, has a poorer rhyme inventory, and must end every clause on some word. The translator faces a contract — which of metre, rhyme, radif and line-for-hemistich to sign, in what register — and, having signed, what to pay with. Each clause is paid in a different coin — content words of the wrong shape, the line-end slot, word order — and the record prices two of them before a line is written and could not price the third.

2. What published translators do

The regularities:

1. A printed formal contract for verse can be kept clause by clause, and where the form is an arrangement rather than a count, the keeping can be audited mechanically, with teeth. Leaf's English holds his own scheme's arrangement at 0.05 violations where chance sits at 0.50. What is measured is lexical-stress alignment with his printed schemes — not the Persian rhythm as anyone hears it — on lines already at the exact count (RS-20260901b-leaf-pattern §2, §7).
2. What defeats a stress language on a quantitative measure is a stress budget, not a want of discipline. In his long-heaviest measures Leaf puts essentially all the stress a line has onto long positions and still cannot fill them; the shortfall is arithmetic on the scheme, knowable before a word is written. His printed mechanism held; his printed list of hard cases did not separate (RS-20260901b-leaf-pattern §5).
3. The one systematic breach of his contract was the one he printed a misgiving about — ode XXI's unaccented -eth rhymes, "without much satisfaction"; and his L-REPEAT self-report held against the project's registered prediction that it would fail (RS-20260830-leaf-contract §3, §5; family K holds the general finding).
4. Line length does not separate the two hands; what does — habit or purchase at the line end — is unmeasured, and after two independent gates is known to sit at the edge of what blind coding can hold steady (~0.80 agreement, never two certified seats). §7.42's instruction — you can, at this price — stands unexplained (RS-20260901-inversion-habit §6; RS-20260903b-line-end-order §4).
5. A rhyme obligation costs the qāfiya slot, by about 5.5×, and the displaced sense goes nowhere in particular. The pre-rhyme concentration (~1.7× uniform) is identical in lines that must rhyme and lines that need not — a fact about the English verse line, not about rhyme (RS-20260902-rhyme-slot §4, §6). On the 14 of 58 renderings that carry an English radif, of the 10 ordinary senses located in such rhyming lines, none stood on the trailing radif words and one stood on the chime itself: the slot a rhyme spends is the rhyme, not the printed last word (RS-20260908-chime-slot §2). Unresolved: recomputed, the radif stratum's effect is +0.157 against +0.241 without a radif, a gap whose interval [−0.128, +0.364] crosses zero on 8 renderings; §7.49's "in both strata" claim is neither confirmed nor reversed.
6. Even a bayt-for-bayt hand moves senses across lines: 57.1% of Payne's located carriers stand outside their own bayt's two English lines, so the record says where a sense-word sits given its line's obligation, never that the rhyme pushed this sense out (RS-20260902-rhyme-slot §7).
7. Enjambment and apparent line-end inversion are confounded by construction: in the lead's frozen gold all 21 items whose clause runs past the couplet are INVERTED, against 21 of 39 complete-clause items (RS-20260903b-line-end-order §5; method note (bsx)).

3. What this project's own practice found

All seven renderings of the five poems are the lead's, frozen before any design or published page was opened, contamination: none (0 shared 7-grams, longest runs 3–4 tokens against the published hand's whole book, which contains that ghazal in four of the five cases — sh115 has no published English rendering in reach, hafez-derakht/dependence-note.md); none is judged for quality.

1. The full contract, signed — T-hafez-bekonad-R57-v1, غزل ۱۸۷ (Leaf's own ode XV), metre 2 at fifteen syllables, two arms differing in the metre clause alone (R57). The radif survives the metre: will do at 8 of 8 rhyming positions — an existence result at the one forbidden cell (family C). Thirteen losses in fourteen lines, twelve of them content words squeezed out by a slot of the wrong metrical shape, none caused by the rhyme or the radif: صد a hundred twice where Hafez repeats it; مشفق → ruth and طبیب → leech (the slot wants a stressed monosyllable — the metre bought two archaisms the hand did not want); فاتحهٔ صبح → the dawn-psalm (a Muslim prayer becomes a Christian one); a root-play. Two consecutive stresses at positions 7–8 filled themselves with compounds — heart's fire, deep dark, high heav'n; internal-judgment-only (RS-20260830-leaf-contract §8).
2. The exhibit of a line end as fixed room. Leaf's ode XV rhymes on -end, this rendering, written unseen, on -ending: opposite purchase — "The bitter pray'r of the midnight a hundred ills shall amend" (Leaf keeps the numeral, drops the radif) against "One prayer at night, in the deep dark, all ills' defending will do" (keeps the device, drops صد). An exhibit, not a measurement (§7.43.4).
3. The hard measure, signed — T-hafez-derakht-R57-v1, غزل ۱۱۵ in metre 6 (sixteen syllables, three longs to a short, one of Leaf's named hard cases): radif آرد as brings at 8 of 8. The registered check that the coder would find at least one violation in these lines failed: zero on 44 short positions across the 11 lines it could read; the lead fills long positions at 0.803 against Leaf's 0.650 in the same measure, and the log names the wall — "I never once produced four consecutive units of three real stresses." (RS-20260901b-leaf-pattern §6).
4. The radif bought by inversion, one line at a time — T-hafez-daasht-R58-v1, غزل ۷۷, radif داشت as had at 9 of 9 in a plain modern register declared in writing before the first line. Reach, tallied from the frozen log: LOCAL 6 · SPREAD 3 · REBUILT 0 — the predicate was never reshaped; the price the log names is the subject–object juncture with no verb between them. One English collocate served all nine complements; the chime HELD 3 of 9, REFUSED 6 (RS-20260901-inversion-habit §5).
5. The line end with nothing to carry — T-hafez-shahed-R60-v1, غزل ۱۵, no radif, same declared register, every line end graded: 20 line ends, 19 CANONICAL, 1 INVERTED, the one where the Persian itself fronts the object; 0 inversions at the 11 rhyming positions. Beside item 4, the same hand inverts where a radif obliges it and almost nowhere else — but one hand who knew the question, 36 line ends, no rate (RS-20260903b-line-end-order §6).
6. The rhyme clause alone — T-hafez-boro-R59-MONO-v1 / -PLAIN-v1, غزل ۳۵, line held fixed (L-LINE), no metre, arms differing only in whether line ends must chime: mean distance of a located sense from its line end 4.13 rhymed, 1.89 unrhymed (Payne on the same ghazal 4.33); the MONO log's own count, frozen first: 4 of 8 rhyme-position senses reached the rhyme word, 4 displaced by one to four words, at three non-positional prices — a dulled chime, one inversion, two verbs made periphrastic. Not evidence: n = 1, and the hand conceived the measure before writing (RS-20260902-rhyme-slot §9).

4. The options

• Sign the whole contract (metre + monorhyme + radif, bayt for bayt; R57). Keeps a form that can be audited as an arrangement, and the radif survives it. Paid in the author's small content words of the wrong shape — numerals, epithets, realia, a root-play — and, at stressed-monosyllable slots, in archaisms and referent changes the hand did not choose (§3 item 1). The measure must first pass the stress budget (regularity 2; deficit 0.23–0.33 of long positions in Leaf's long-heavy measures).
• Release the metre, keep the rhyme (R57 FREE; R59 MONO). The line's length is free but the qāfiya slot is spent: an ordinary sense reaches it at 0.049 against 0.270 when free, and the displaced sense lands nowhere plannable (regularity 5).
• Release the rhyme, keep the line (R59 PLAIN). The line ends where the sense ends; the pre-rhyme concentration remains, because it is English's; what is lost is the chime and, per the one log, the inversions and periphrases the chime forced (§3 item 6).
• Carry the radif by line-end inversion in a plain register (R58). Buyable one line at a time — verb alone moved at 6 of 9, predicate reshaped at 0 (§3 item 4); the price (family C, §7.42) is well-formedness at about 2.4 of 10 in plain lexis, 1.9 in archaic, no single figure (0 to 6), archaism not discounting it.
• What the option set depends on: whether a radif exists and its part of speech (family C); whether the source's rhyme senses lie inside one English rhyme family (family B); the measure's long-fraction; the declared register (it does not change the inversion's price); and what a reader takes from any of it — unmeasured, Tier D exhausted.

5. Guidance

For a translator

1. Before signing a measure, do the stress arithmetic. Transcribe the scheme, count its long positions per line against the stresses a line of your English can carry; the shortfall is a budget you cannot fill however you place stress, rising with the measure's long-fraction. — evidenced (FA→EN, one hand's whole book, 18 measures; one lead poem in a 3:1 measure).
2. If you sign, decide before the first bayt which of the author's small words you will lose — numerals, epithets, realia names — because that is where a fixed measure is paid, not in the rhyme or the radif; and watch the slots that forced an archaism and a change of religion. — evidenced (FA→EN, one hand, one poem, frozen log; internal-judgment-only).
3. Count the rhyme slot as already spent. A sense that would naturally close the line does so about a quarter of the time when the line is free and about one time in twenty when it must rhyme; do not plan on the word before the rhyme — it takes the displaced sense no more often than in an unrhymed line; plan where the line is still free. — evidenced (FA→EN, two published hands, 747 sense positions).
4. Where an English radif trails the chime, protect the chime, not the printed last word: of the 10 ordinary senses located in such rhyming lines, none stood on the trailing radif words; one stood on the chime. — evidenced (FA→EN, 14 renderings, 10 items — small; the by-radif stratum comparison is unresolved).
5. A finite-verb radif can be bought by inversion one line at a time in plain modern English — expect the verb alone to move at most positions, a second constituent at some, the predicate at none; expect the hardest reading at a subject–object juncture with no verb between; and do not expect a register, archaic or plain, to change the price; declare the register in writing first (family C). — evidenced (FA→EN, one hand, one poem; price from family C).
6. Do not infer from a published hand's inversion at the radif that he inverts everywhere, or the reverse. The project could not measure it on two hands; its own hand inverted 9 of 9 where obliged and 1 of 20 where not. — untested as a rule about hands; one hand's record only.
7. Where the form is an arrangement, audit your own lines against it mechanically on lines that already carry the count — the audit separates a hand who tried (0.05) from chance (0.50) — and read the result as lexical-stress alignment, never as the rhythm a reader hears. — evidenced (FA→EN, one hand's book; one lead poem).

For a pipeline

1. Enumerate: the source's measure as a long/short scheme (from the rukn names, checked against a second attribution — Ganjoor matched Leaf's table at 20 of 20 ode rows), the rhyming positions, the radif and its part of speech (family C), and the sense at each hemistich end. — untested as a step; every component has been run by hand.
2. Compute the stress budget per target line: long positions against the stresses the line can carry; flag measures above about 2.5 longs to a short. — untested as a step; the instrument (scan.py) exists.
3. Set the contract parameters: metre on/off; rhyme on/off; radif on/off; line-for-hemistich on/off; register, declared as a written list of excluded forms. — untested.
4. Render and freeze the log before any comparison, recording per rhyming position the inversion's reach (LOCAL / SPREAD / REBUILT / REFUSED) and the chime (HELD / SHORT / REFUSED), per line the order grade with an enjambment flag, per bayt the losses and alternatives. — each component evidenced (FA→EN, one lead poem: losses R57, reach and chime R58, order grade with enjambment flag R60); the combined format and the autonomous step untested.
5. Check: (a) stress-arrangement audit against the scheme on exact-count lines; (b) for each hemistich-end sense, where it landed — distance to the line end and to the qāfiya slot; (c) the per-bayt loss list. Do not report a line-end word-order rate: no design has certified two seats on it, and any count must first separate enjambment from inversion (note (bsx)). — untested.

Human entry points. Step 3 — which clauses to sign — is a purpose decision (declared-purpose, wiki/goodness-senses.md) and routes to a person; without one, default to the best-evidenced cheap contract: rhyme and radif where family C says the radif is carriable, metre released, plain register declared. Step 5(c)'s loss list is where a person may overrule a referent-changing loss (a dawn-psalm); unreviewed, it stands and the log says so.

Application. Not applied: written at close-out without a translation limb.

6. Not evidenced, and open

• One author, one pair, two published hands, one unaudited for metre (Payne's OCR is unusable; his title-page promise of "the original forms" was never checked). No other verse tradition, no period control corpus of English verse, no human reader or scansion — Tier D is exhausted.
• The metre audit reads lexical stress, not rhetorical stress, conditional on the 301 exact-count lines (86 resolvable lines excluded, 73 of them off by one syllable).
• No inversion rate exists for any published hand; the local-purchase reading rests on one hand at 36 line ends.
• The rhyme-slot effect is observational — nothing assigned; the detector misses about a quarter of rhyming lines (so 0.2207 is a floor); senses migrate; the lead pair is n = 1.
• Open: whether the rhyme's cost differs with and without a radif — the point RS-20260908-chime-slot §4 leaves undecided; needs more than the 14 radif renderings in hand. Open: whether the local-purchase reading (§3 item 5) holds for a published hand — needs a two-seat-certified coding or an independent human coder, unbuilt (RS-20260903b-line-end-order §8). Open: the end-weight conjecture on the inversion's price (§7.42.4, family C) and the rhyme-family conjecture (§7.44, family B, not refuted) — named in this family's logs, owned elsewhere. Open: whether consecutive-stress positions generally draw English compounds (§3 item 1; one hand). Open: a matched-period verse control for the stress audit (RS-20260901b-leaf-pattern §7).

7. Sources consumed

• v0.2 §7.43 (S233, RS-20260830-leaf-contract, A-leaf-hafez): the five-clause audit, the correction of §7.41.3, the exhibit of §7.43.4, T-hafez-bekonad-R57-v1's loss table; §7.43.2 is family K's.
• v0.2 §7.47 (S237, RS-20260901-inversion-habit): the ROOM account contradicted; the first word-order withholding; T-hafez-daasht-R58-v1.
• v0.2 §7.48 (S238, RS-20260901b-leaf-pattern): the arrangement audit, the stress budget, T-hafez-derakht-R57-v1.
• v0.2 §7.49, line-end half (S239, RS-20260902-rhyme-slot), corrected by RS-20260908-chime-slot (S254; no v0.2 section — v0.2 is frozen); the rhyme half is family B's.
• v0.2 §7.51 (S242, RS-20260903b-line-end-order): the second withholding, the enjambment confound, T-hafez-shahed-R60-v1.
• Sections the row's titles hid, consumed by other families and cited here only where they price a line end: §7.42 (family C, the inversion's price); §7.44 (family B, the pre-rhyme clause §7.49 withdraws); §7.41.3–§7.41.4 (family C); §7.26 and §7.32 (family B, line-end position effects). Also hidden: T-hafez-amad-R57-v1 (S234, R57 FREE arm, RS-20260830b-rhyme-family's limb, family B). Regimes R57–R60; the five translation pages of §3.
• Withdrawn along the way: §7.41.3's "Leaf, writing plainer, never uses [the inversion] and drops every one of those radifs" — withdrawn by §7.43.1 (RS-20260830-leaf-contract §7; ode XVI inverts 9 of 9). §7.41.3's register clause — withdrawn by §7.42 (RS-20260829b-inversion-price, family C). The ROOM account — contradicted before it was ever recommended, RS-20260901-inversion-habit §3. §7.44.1's "they sit in the word immediately before the rhyme" and §7.44.4's second sentence — withdrawn by §7.49.2, replaced by §7.49.4 (RS-20260902-rhyme-slot §6). §7.49.1's "a factor of eleven" — corrected to about 5.5 by RS-20260908-chime-slot §3. §7.49.1's "in both strata" robustness claim (F3) — weakened to unresolved by RS-20260908-chime-slot §4. T-hafez-bekonad-R57-v1 log §D2's account of its own rhyme family — corrected by §7.44.2 (family B); the frozen log stands. Registered predictions that failed, recorded not withdrawn: L-RHYME ≥ 24 of 28 at 1.000 (RS-20260830-leaf-contract §3); L-REPEAT would fail (§5); the coder would find ≥ 1 violation in the lead's metre-6 lines (RS-20260901b §6). Withdrawn before any figure: the FILL test (RS-20260901b §5) and the R60 limb's "upper bound" reading (RS-20260903b §6).