In the preface to Antonio Machado: Selected Poems, the translator Alan S Trueblood (what a gift of a name!) writes:
"One cannot hold today that a poet's voice in translation should sound as if he had been writing in English all along. ... Some aura of foreignness, individually and culturally marked, should survive re-creation."
By slight contrast, Rob Mackenzie, concludes the first of his reflections on translating Quasimodo by saying:
"Each translation must stand as a poem. The translator can only hope against hope that he has communicated something of what was in the original writer’s mind."
These statements could be taken as describing two separate approaches to the art of translating poetry. Trueblood's emphasis strikes me as bearing the mark of literary or academic translation as an act in itself, in that it is separated from any other creative work. By this I don't mean to deny that such translation is a creative act, because such a denial would be absurdly naive. I simply mean that the literary/academic translator's primary motivation is not necessarily to make a poem really work in the target language.
On the other hand, Rob's concern seems that of a poet: to create a poem in English that really works but still carries something of the imagery and impact of the original. This endeavour cannot be separated from the translating poet's work on their own poetry, because something of their voice will most likely leak into their translations and, if the translating is carried out attentively, something of the translated poet's voice might well inf(l)ect the translator's own writing. The effectiveness of this can be seen in Ted Hughes's translations of Yehuda Amichai.
Perhaps this is where the notional difference between a translation and a version (see this post) comes into play. If so, does that mean that the creation of a version is the proper concern of a poet and the creation of a translation the proper concern of a professional translator? It's not a thought I accept, since it seems an unhelpful narrowing of the fields. Still, it might throw some light on why some people feel uneasy about so-called versions.
The two approaches, of course, both carry undeniable benefits and disbenefits. In my handful of translations from Rilke into Scots (there's one here) or English, I think I've taken what I've described as the poet's approach. This is partly because of the failure of translations I'd come across to show the reader anything much of what drew me to Rilke.
Perhaps the poet's approach is, in the long run, personally less rewarding than the academic translation because it's pretty much doomed to failure. Eliot Weinberger says in "Translation", the second of his "3 Notes on Poetry" (published in Outside Stories):
"A translation that sounds like a poem in English is usually a bad translation.
"A translation that strives for the accuracy of a bilingual dictionary is always a bad translation.
"A translation must sound like a translation written in living English".
By contrast, it strikes me that nobody would accept a translated novel that did not read as a good novel. Basically, a good translation is a tall order for a translator taking either approach.
On reflection, it surprises me that translation has not been given greater prominence in Scottish literature. After all, as a three-voiced country, we're always translating between ourselves, let alone from other cultures. Perhaps we need to find a vocal and visible way of celebrating our translators.
I'll leave you with another thought from Weinberger's essay:
"Translation theory, however beautiful, is useless for translating. There are the laws of thermodynamics, and there is cooking."