Repository path: workshop/translations/hafez-derakht/R57-v1/translation.md · rendered 2026-09-09

T-hafez-derakht-R57-v1 — Hafez غزل ۱۱۵ under Leaf's contract, in the measure Leaf called impossible

Source: workshop/translations/hafez-derakht/source.md — Ganjoor, Qazvini–Ghani, sh115, 7 bayts, radif آرد, qāfiya in -ār, metre هزج مثمن سالم = Leaf's metre 6. Regime: R57 (Leaf 1898's printed contract), clauses 2–5. Translated by: the lead, in session, at no API cost. Contamination: none, measured — see ../dependence-note.md. Lineage: no prior lead rendering of this poem. No published English of it was opened, before the freeze or after.

One arm only, and the reason is declared. R57 clause 6 offers a paired LEAF/FREE design. It is not invoked here. The contrast this session needs is against Leaf's own metre-6 ode, which exists and is on the shelf; a second lead arm would only compare the lead with itself.

Tier D is NOT PASSED. Nothing below is a quality claim, and the lead does not judge its own translation (charter §5).

The measure, stated before the first line was written

Metre 6 of Leaf's table, hazaj-i musamman-i sālim, مفاعیلن four times:

 ⏑ – – –   ⏑ – – –   ⏑ – – –   ⏑ – – –
 x / / /   x / / /   x / / /   x / / /

Sixteen syllables. Twelve long, four short. Three longs to every short — the ratio Leaf names in his Introduction as the case where "the English language will not supply stress enough."

Quantity is rendered by stress, in Leaf's own terms and with his own asymmetry, which R57 clause 2 takes from him and which is the whole difficulty:

• a stressed syllable must stand at a long position — "I have therefore accepted all stressed syllables as long";
• a long position need not carry a stress — "It is not possible to treat all unstressed syllables as short … in a large majority the quantity is uncertain."

So positions 1, 5, 9 and 13 of every line are barred to stress, and the other twelve are to be filled with stress as far as English will go. Leaf's own worked example of the second permission is in his footnote to ode IV: he prefers "Bokhara town and Samarcand" to the "quantitatively exact" alternative.

The rendering

Implant love's tree: it all thy heart has longed for, home in full gain brings; Uproot hate's shoot: that plant, left green, beyond all count its own pain brings.

To those wild hearts that throng wine's door be thou all grace, for thou art guest: The head's dull ache is thine, O soul, if thy wine's mirth its own bane brings.

The night of talk as spoil hold fast, for when our time is past and gone The wheel turns on a thousand turns, and day and night in long train brings.

Thou driver of the litter where the moon's own cradle owns thy hand, For God's sake plead Majnun's sore need, and bid that hand her lone wain brings.

Demand life's spring, my heart; else here this lawn each year renews its store: As wild-rose bears a hundred blooms, the bulbul's thousandfold strain brings.

For God's own sake, my torn heart made a pact long since with thy dark hair — Command that mouth of sweet red stone that heart of mine to swift rein brings.

Within this close, the grey-head Hafiz asks one boon of God at last: To sit hard by a brook's bright lip, and there one cypress in chain brings.

Radif: brings, at all eight rhyming positions, rendering آرد — a transitive finite verb, the cell framework/v0.2 §7.36.1 calls unreachable. Qāfiya: eight distinct -ain rhymes immediately before it — gain · pain · bane · train · wain · strain · rein · chain.

The scansion, written out by hand so it can be checked rather than believed

x = a syllable the hand intends unstressed, / = stressed. Target, every line: x / / / x / / / x / / / x / / /. A • marks a licence — a long position the hand could not fill with stress, which Leaf's contract permits. A ✗ would mark a violation — a stress at a short position, which it does not.

By the hand's own count, over 14 lines and 224 positions:

• Short positions kept free of stress: 56 of 56 — 1.000. No violation of the clause Leaf actually signed.
• Long positions carrying a stress: 137 of 168 — 0.815. Thirty-one uses of the licence.
• One line, 1b, holds all sixteen positions. No other does.

These are the hand's intentions, not a measurement: an independent coder reads the same lines in E-20260901b-leaf-pattern, and where it disagrees with this table the disagreement is reported and not resolved in the hand's favour.

Translator's log

Frozen 2026-09-01, before the study limb was designed and before any coder was run over these lines. Decisions and alternatives; not quality claims.

The thing that governed everything: where the four short syllables fall

The measure gave me four free slots per line — positions 1, 5, 9, 13 — and twelve slots I had to try to fill with stress. English gives the opposite. Its function words are unstressed and it has a great many of them; its content words are stressed and three in a row is already a strain ("deep dark wood"). So the composition method that emerged, and I did not have it at the start, was:

build each line as four units of [one unstressed syllable] + [three stressed], and put at positions 1, 5, 9 and 13 either a function word or the unstressed syllable of a polysyllable.

Two devices did nearly all the work and both were found by trial:

1. Unstressed derivational prefixes at position 1. im-PLANT, up-ROOT, de-MAND, com-MAND, with-IN. Six of the fourteen lines open on one. My first drafts of 1a and 1b opened "So plant love's tree" and "Then rend hate's shoot" — and so and then, read naturally, take a stress, which is a violation at position 1 and not a licence. Implant and Uproot fixed both lines at once and are closer to the Persian imperatives («بنشان»، «برکن») than plant and rend were.
2. Trisyllables and disyllables straddling the boundary at 8–9 and 12–13. HA-fiz in 7a sits on 8–9 exactly; CRAD-le in 4a on 12–13; CY-press in 7b on 12–13; THOU-sand-fold in 5b on 12–14. Where I could not find one, position 9 forced a function word, and 9 was the hardest position in the line — it is the one that has no natural syntactic break before it.

What I could not do at all. I never once produced four consecutive units of three real stresses. The licence — Leaf's permission to let an unstressed syllable stand at a long position — was used thirty-one times in fourteen lines, and I could not have written a single line without it. Line 1b is the only one that holds all sixteen positions, and it does so because "that plant, left green, beyond all count" happens to be four content monosyllables and a prefixed adverb in a row. I could not repeat that.

Where the licences cluster is not random and I noticed it while writing. They fall overwhelmingly at positions 3, 7, 11 and 15 — the second long of each three-long run. Of the 31, nineteen are at those four positions. English will give you x / / readily ("a brook's bright"); it will not give you x / / / without a compound or a run of monosyllabic nouns. The third long is what English cannot supply, and the second is what it gives up first when the sentence needs a preposition or a possessive.

Bayt by bayt: what was live, what was chosen, what was lost

1a–1b. The pair is a matched imperative in Persian («بنشان» / «برکن»), and the parallel survived intact — the only bayt where I got the Persian's symmetry for free. «کام دل» "the heart's desire" became "all thy heart has longed for", which is longer than the Persian and buys the syllables position 9–11 needed. Lost: «به بار آرد» is literally "brings to fruit", and the horticultural metaphor that ties the verb back to the planted tree is gone; "home in full gain brings" is mercantile where Hafez is arboreal. Alternative considered and dropped: "to full fruit brings", which keeps the metaphor and breaks the -ain rhyme family.

2a. «رندان» rinds — the tavern libertines — has no English word. Live options were the loanword rinds (unglossed, and Leaf's own book uses topers), libertines (four syllables, wrong register), and a periphrasis. I chose "those wild hearts that throng wine's door". The loss is real: rind in Hafez is a specific spiritual position, the holy scapegrace, and "wild hearts" makes it merely riotous. «خرابات» likewise: the kharābāt is a ruin and a wine-house at once, and "wine's door" keeps only the second.

2b. «خمار» khumār is the morning after, and English has only hangover, which cannot go in this register, or ache. I split it: "the head's dull ache" at the front of the line and "its own bane" at the rhyme, so the one Persian noun becomes two English phrases. «جانا» — the vocative "O soul" — survived at position 7, which is a long, and it is one of the few places a stressed monosyllable arrived free.

3a–3b. «غنیمت دان» is "know it as spoil, as war-booty", and "as spoil hold fast" keeps the plunder. What I lost is the «لیل و نهار» pair, which is Arabic in a Persian line and marks a register lift; "day and night" is flat English and reads as idiom, not as citation. I could find no way to mark it. «گردون» — the turning sphere — kept its turning: "the wheel turns on a thousand turns" is a figura etymologica the Persian also has («گردش کند گردون»).

4a–4b. The hardest bayt. Persian has «عماری‌دار», one word for the man who drives the camel-litter, and English has none; "Thou driver of the litter" takes four syllables to say it and pushes the moon's cradle to the end. Majnun stayed and the phrase «گذار آرد» went. The Persian asks the driver to bring his passing by Majnun — to route the caravan past the madman in the desert — and the radif brings had to take an object that rhymed in -ain. I could not get "past Majnun" to the end of the line under the measure; every arrangement that did put a stress at position 9 or 13. So the object became "her lone wain", the litter itself, and the direction of travel is only implied. That is the largest single loss in the poem, and it is a loss the metre caused: I had the words and could not place them.

Majnun also sits at positions 5–6 precisely because maj- is unstressed, which is luck.

5a–5b. The pun is gone. hazār — "a thousand" and the nightingale — cannot be carried, and this is the third time this project has met that particular impossibility. What I did instead was put the number and the bird in the same word: "the bulbul's thousandfold strain", so the thousand attaches to the bird's song rather than to a count of roses. That is a compensation and it is not the pun; a reader who does not know Persian gets one joke fewer, and cannot know it. «نسرین» is the eglantine or dog-rose; "wild-rose" is right and loses the name.

6a–6b. «قرار» twice, as compact and as rest. I got pact (6a) and rein (6b), which are two English words, so the identity is broken — but rein at least keeps a settling-down and sits in the rhyme family. Alternatives were rest (breaks the rhyme) and chain (which I needed for 7b). «لعل نوشین» — the honeyed ruby, the lip — became "that mouth of sweet red stone", four syllables longer than the Persian and, I think, harder: the Persian names the stone and lets the mouth be understood, and English cannot leave the mouth out and still be read.

7a–7b. The maqta' keeps the takhallus, and "the grey-head Hafiz" fits the measure exactly because HA-fiz is a trochee and lands on 8–9. «کنار» is side and embrace at once; "in chain" takes the embrace and drops the riverbank, and "hard by a brook's bright lip" takes the bank a line earlier. The cypress stayed a cypress and is therefore, for an English reader, only a tree.

One thing I want on the record because it surprised me

The rhyme was easy and the metre was not. Eight -ain monosyllables arrived without strain, and the radif brings — a transitive finite verb in the position framework/v0.2 §7.36.1 says cannot be filled — cost me nothing at all after the first bayt, because once the object has to precede the verb in every line, the inversion stops being a decision and becomes the sentence pattern. What cost me the whole morning was twelve long positions and four short ones. If this rendering is worth anything as evidence, it is evidence about the measure and not about the rhyme.