Repository path: workshop/translations/saadi-ghazals-radif/R51-v1/translation.md · rendered 2026-09-09

Six Sa'di ghazals — English under R51

Source. سعدی, «غزلیات» ۱۶, ۳۱, ۴۹, ۵۰, ۵۸, ۷۰, whole. Copy-text, its single-witness limit, and the rhyming structure frozen from the Persian before any English existed: ../source-frozen.md. 39 bayts, 45 rhyming positions.

Regime. R51 (workshop/regimes/R51-radif.md). The rendering of record is AB: at every rhyming position the English line ends in the same word or phrase (the radif), and the word immediately before it rhymes with the corresponding word at every other position (the qāfiya).

Contamination: suspected. Declared before any English was written, on the standing ground that Sa'di is a much-translated poet and English Gulistans and selections of the odes are certainly in the lead's training data. The measurement is in §8; it was run after this log was frozen and committed, and it locates runs rather than licensing an independence claim. Priming declared: the lead has not knowingly read an English rendering of any of these six ghazals, and no published hand was opened for any of them before the freeze. That is a statement about what the lead can recall being told, not a measurement, which is why §8 exists.

Cost: $0. Lead translation, never ledgered (charter §3, A4).

1. What the English radif has to be, and why the choice is forced before the first line

A Persian rhyming position is […qāfiya][radif], and nothing intervenes. So an English hand choosing a rendering for the radif is simultaneously choosing what part of speech may stand at the end of every clause in the poem, because whatever word ends the English radif must be preceded by a word that can be rhymed on. That decision is made once, for the whole poem, before any line.

Persian is verb-final. English is not. This turns out to be the whole story, and the enumeration below was written before any of the six was rendered.

The reasoning, in one sentence. A Persian radif can be any word at all because Persian puts the verb last and a Persian clause therefore ends wherever the poet likes; an English clause ends on an object, a complement or an adverbial, so an English radif can only be something an English clause can end on — and the qāfiya slot immediately before it inherits that constraint.

2. ۴۹ — radif آنجاست "is there"; English radif is there, qāfiya family ‑EST

Enumeration for the qāfiya family, written before the lines (rule 1). The eight positions must name: a resting-place · the cure of a sick heart · a rogue who has stolen the heart · a wandering star · the beloved · a confidant of secrets · the holder of the heart · the lodging of free men. Families enumerated: ‑EST (rest, breast, best, nest, guest, quest, confessed, possessed, dispossessed, oppressed, blessed) — reaches all eight; ‑ART (heart, part, art, mart, smart) — reaches three; ‑ĪR (here, dear, near, sphere, seer) — reaches four but collides with there. ‑EST chosen.

Blessed that ground, where my love's own rest is there; the soul's whole ease, and the cure of the sick breast, is there.

Here in this place I am only a shape with the life gone out; my heart is off where that rogue who has me dispossessed is there.

My body is here, and sick; my heart is there, and lodged; the sphere is here — but the star that will not rest is there.

Come at last, morning wind: if you carry a scent at all, turn by Shiraz, for the one I love best is there.

To whom shall I tell the heart's ache? With whom eat the heart's grief? I will go where the keeper of all I have confessed is there.

My heart does not lean to the show the meadow makes: the show the heart wants is where the one who has it possessed is there.

Sa'di, what will you do in this broken lodging? It is not your place. Strike your tent — where free men have their nest is there.

Bearers: rest · breast · dispossessed · rest · best · confessed · possessed · nest. Seven distinct at eight positions — the same 7-of-8 Sa'di has, and the repetition falls where his does not, at 1a/3b rather than 1a/4b.

3. ۵۰ — radif برخاست "arose"; English radif arose, qāfiya family ‑ĀME

Enumeration. Positions must name: reason's reproach · the decree of safety · the road of blame · the dust of regret · the name of chastity · the cypress's forfeit · the pine's stature · the Resurrection. Families: ‑ĀME (blame, claim, shame, name, fame, frame, lame, aim, same) — reaches all eight, three of them exactly (blame = ملامت, name = نام, frame ≈ قامت); ‑ŌZE (rose, chose, prose, repose, throes) — reaches two; ‑EST — reaches three. ‑ĀME chosen.

I made love my business, and against me my reason in blame arose; for out of the man who has loved, safe-conduct's whole claim arose.

Whoever has sat in a private room with a rose-faced beauty never yet from the highroad of blame arose.

Who have you heard of that spurred the roan of love's grief and behind whose heels no dust of shame arose?

Love won; and out of the corner-sitting devout the name of the modest, the credit of holiness, their fame, arose.

In a garden where that laughing rose-bush took its seat, the free cypress on one foot, in shame, arose.

With what splendour the hundred-petalled rose came open I do not know, or in what height and stature the pine's own frame arose.

Yesterday she sat a while by Sa'di, all ceremony: the riot sat down; when she rose, the world's last claim arose.

Bearers: blame · claim · blame · shame · fame · shame · frame · claim. Five distinct at eight positions, against Sa'di's seven. The shortfall is the finding, not an accident: see §6.

Two positions carry a strain and it is declared here, not discovered later. At 1b and 4b Sa'di's برخاست means departed — the decree of safety left the lover; the name of chastity left the devout — and English arose does not mean that. A radif is one form in Persian and one word in English, and Sa'di uses this one in two senses inside seven bayts. The English hand must pick one and pay at the others. Marked SENSE-STRAIN in §6.

4. ۵۸ — radif افتاده‌ست "has fallen"; English radif has fallen, qāfiya family ‑ĀIN

Enumeration. Positions must name: someone's lane · the slain · a cage · a breath · honey with a fly in it · a snare of desire · someone's polo-stick. Families: ‑ĀIN (lane, slain, chain, pain, vain, plain, again, remain, domain) — reaches two exactly (lane = کوی, slain = کشته) and three more only with a supplied word; ‑ĀY (way, prey, play, stray, astray, today) — reaches four with one supply; ‑ĒR/‑ĀIR (snare, air, lair, ensnare) — reaches two. ‑ĀIN chosen, on the two exact hits, with the outcome recorded rather than repaired.

My lot at the head of a certain man's lane has fallen, and in that lane many a one like me, slain, has fallen.

Carry my news to the birds of the meadow: that one who sang with you into a cage — and a chain — has fallen.

Tell the heart's-ease, O breath of the dawn wind, that our case, like the daybreak's, on one breath's short reign has fallen.

What can the fetter on the foot do but bear it? Honey it is, and into the honey a fly, in the main, has fallen.

Nobody faults us for playing at wanting — nobody, that is, but the man who to some desire's own chain has fallen.

Sa'di, the state of a man who talks scattered is known to the man who all his life to somebody's mallet and plain has fallen.

Bearers: lane · slain · chain · reign · main · chain · plain. Five distinct at seven positions. Four of the seven are bought with a word the Persian has not got — chain for a cage, reign for the span of a breath, in the main for nothing at all, plain for a polo-ground — and by R44's sense clause, carried into R51 rule 1, a rhyme bought that way is a refusal, not a rhyme. They are marked REFUSED in §6 and the poem is graded SHORT. This rendering is filed as it stands and is not repaired, because the point of grading a regime is to find out what it costs.

5. The three the English radif could not be built for

۳۱ — radif انداخت "cast" — UNREACHABLE

Sa'di's verb is transitive, and its object is different in every bayt: turmoil, blood, talk of you, tumult, me, a glance, his life. Persian puts the object before the verb, so the verb ends the clause and the object supplies the rhyme two words back. English puts the object after the verb. Three English frames were enumerated and all three fail at the same place:

1. …[object] has cast — the word before the radif is the auxiliary has, which cannot be rhymed on, and the subject must then be fronted into a construction English does not have.
2. …[X] was cast (passive) — the word before the radif is was.
3. …[subject] cast (inversion) — the word before the radif is now the subject, and Sa'di's subjects are your beauty, the calamity of your glance, fate, your cypress stature, the enemy, Sa'di. Nothing rhymes with anything, and nothing can be made to without replacing what the poem is about.

So the poem is rendered without the radif, at no rhyming position — and printed here as what the hand actually produced, because a translation of record that quietly omits a whole poem is worse than one that prints its failure:

What a turmoil your beauty has cast upon the world — that one cannot cast a look off you for a breath.

The calamity of your unkind, blood-drinking glance: what blood it has cast into the hearts of kind friends.

I was left on the far side of reason and of my own safety the day fate cast talk of you into the middle of things.

Neither garden is left nor orchard: your cypress stature came up, and cast an uproar through garden and orchard.

Be a friend, and do not cast me out of your eye — the enemy has cast me on people's tongues for your sake.

By your eyes! An eye that is taken off you would be wasted, cast up at the moon of the sky.

This very story will come one day to my friends: that Sa'di went after the beloved and cast his life away.

۱۶ — radif مرا "me / to me / my" — UNREACHABLE

مرا is a single oblique form doing three jobs, and it changes job from bayt to bayt: possessive at 1a (my senseless heart), 1b (my liver), 2b (my mind); dative at 4b (to me), 5b (to me); accusative at 3b (makes me forget) and 6b (do not sell me). English has no one word that can stand at the end of a clause in all three roles. Where English me can end the clause at all, the word before it must be a transitive verb — which exists only at 3b and 6b — and those two verbs (forsake, sell) belong to no common rhyme family. Two positions of seven can host a bearer, and two bearers cannot chime.

So long as the load of your grief lies on the stunned heart in me, your love's own burning will never settle the seething in me.

No thought of the rose or the hyacinth crosses my mind while that lock and that bright ear-side keep their hold in me.

A draught more bitter than the poison of parting from you is needed to make the sweetness of your nearness forsake me.

Every night my head goes down on the pillow with the grief of your going — if never a day put your arms about me.

If they hand me a hundred cups of nectar and not your mouth, then by your mouth: it is poison that nectar brings me.

Sa'di, in the fist of the headsman of grief for you, says: "I am your slave, your slave. Give me to the killing, but do not sell me."

The radif itself is held at all seven positions — every line ends in me. Only the chime is unreachable. That is the cleanest instance in the six of the two halves of the device coming apart, and it came apart by itself.

۷۰ — radif ترست "is more —er" — SHORT

The radif is a comparative suffix plus a copula, so every bearer in the Persian is an adjective stem and the radif does the comparing. English can imitate this exactly — brighter is, lighter is — because English also builds its comparative on a suffix, and the rhyme then falls on the stem exactly as Sa'di's does. What English does not have is a family of adjectives that can mean, in turn: nimbler, more joyous, more fearless, apter, purer, more torn, more dust-like. Four of the seven come out on ‑IGHTER; three do not.

O you, than the walking cypress whose stature the lighter is; the heart, at your face, than your face itself the brighter is.

I have stopped being afraid of death's blood-drinking blade, since beside your blood-spilling glance it is not the bolder.

The dress of meaning has always sat well on me; on your straight body, though, it the tighter is.

If an enemy taunts my clean look at you, thank God, my friend's own skirt is the whiter — is.

Since the rose of your face came open in beauty's garden my patience's curtain is torn worse than the rose's hem.

Set your foot on Sa'di's eye, if you will walk: a hundred degrees below the dust of your door is dust.

Four positions held, three refused, and the four hold on an inversion — "the lighter is" for "is the lighter" — which is available in English verse but is archaic in a way the Persian is not. That is a cost of the frame, not of the rhyme, and it is declared here because it travels into any window drawn from this poem. The three refused bearers would have had to mean more dauntless, more torn and more dust-like, and no ‑IGHTER word means any of them.

6. Grading — the hand's reading, and the tool's

The hand's grade per rhyming position, written before tools/rhyme_pairs.py was run:

The mechanical grade is in grading.json, produced by tools/rhyme_pairs.py over the declared bearer lists, and both readings stand where they differ (R51 rule 1). The tool's STRICT test requires the phoneme immediately before the rime to differ between the two bearers, which is the Persian prosodists' own شایگان objection and is stricter than what an English poet counts as a full rhyme: it rejects rest/breast and shame/fame, which the hand counts as rhymes. Where the two disagree, the disagreement is reported and neither is overridden.

7. The instrument material — RH, RD, NN at the qualifying windows

Declared before the windows were drawn (R51 rule 3): a window is any three rhyming positions from one ghazal whose AB bearers chime and whose radif is held. The three bayts are printed together as a six-line unit. They need not be consecutive: the ghazal is a form of semantically disjunctive bayts — anthologies have printed single bayts of Sa'di for seven centuries — so three bayts of one ghazal set side by side is the form's own reading unit, not a mutilation of it.

The windows and the three variants are in arms.md, written at this sitting, from the same Persian, before any evaluation of them existed.

8. Contamination measurement

Run after this log was frozen and committed: ../dependence.json, produced by tools/dependence_check.py against Sir Lucas White King's The Odes of Sheikh Muslihud-Din Sa'di Shirazi (Kaviani Press, Berlin, 1926; public domain, full text at archive.org/details/odesofsheikhsadishirazi), the one freely reachable English of Sa'di's odes. King's book renders the بدایع and is numbered by its own order, not Ganjoor's, so alignment is by content, done mechanically after the freeze; the result and its coverage are in §8 of the result page and on ../dependence-note.md. Where no aligned comparator exists for a poem, that is said rather than glossed.

Translator's log

Frozen 2026-08-26, before any evaluation of this translation was designed.

(a) The choice that is made before the first line, and is not a craft choice. In every previous verse rendering this project has made, the rhyme word was chosen last: the line is built and the final word is found. Under a radif that order inverts. Choosing the English radif fixes what part of speech may end every clause, and therefore fixes the whole enumeration of possible qāfiya families, before a single line of the poem is written. On ۴۹ the choice took four minutes and the poem took an hour; on ۳۱ the choice took an hour and the poem was never written, because there was no choice to make. internal-judgment-only: this is the largest structural difference between rendering a rhymed form and rendering a radif form that the hand has met.

(b) What actually blocks ۳۱, stated as a fact about the two languages rather than about me. The radif انداخت is a transitive verb. Persian's object-verb order lets a transitive verb end a clause with a rhymeable noun two words back; English's verb-object order does not, and the three English repairs (perfect, passive, inversion) all put an auxiliary or an un-rhymeable subject in the qāfiya slot. Nothing about vocabulary size is involved. The prediction table in §1 was written before any of the six was attempted and it came out 6 of 6 — which is worth recording because it means the constraint is derivable from the grammar of the radif alone and a translator can know before starting whether the shape is available.

(c) Where I paid, and in what coin. ۵۰ is the interesting one. Every position holds, and two of them hold on a word that means the wrong thing: Sa'di's برخاست departed at 1b and 4b, my arose. I could have chosen an English radif meaning departed and paid at the other six instead. There is no third option, because the radif is one word by definition. A radif is a repetition of form, and Persian lets its sense move under the repetition; English repetition drags the sense along with it. internal-judgment-only.

(d) The rhyme family, counted rather than felt. ۴۹ came out with seven distinct bearers at eight positions — exactly Sa'di's ratio. ۵۰ came out with five distinct at eight, and ۵۸ with five at seven, where Sa'di has seven and six. So the standing finding (RS-20260823-run-depth: the binding constraint on a deep English run is the size of the rhyme family) reproduces here, and the frozen source page records the thing that was not expected — Sa'di repeats his own bearers in four of these six poems. Persian faces a version of the same constraint and pays it the same way.

(e) The one place the device came apart by itself. ۱۶ holds the radif at 7 of 7 and the chime at 0 of 7 — a pure RD arm, produced by the grammar of Persian's oblique pronoun and not by any decision of mine. I noticed this while rendering bayt 4 and it is the reason §7's window rule was written the way it was.

(f) What I did not do. I did not go back and repair ۵۸'s four bought rhymes, and the temptation was strong, because three of the four are one word away from honest. R51 rule 1 forbids revising a rendering to move a grade, and the four are the only measurement in this artifact of what the sense clause actually costs at depth 7.