In the shower today I was thinking: I don't know Arabic, have never travelled to the Middle East or North Africa.
I am an American, with a pretty fine-toothed knowledge of American poetry, culture, and jazz.
So when I talk about Lorca's influence in the US, I still miss things if I don't research well enough, or if something simply never comes to my attention.
If I find in 20 minutes on the internet that Lorca is a big deal in modern Arabic poetry, and gather a few references, I pretty much don't know what I'm talking about. Everything comes to me in translation, and even someone with an in-depth knowledge of those milieus could miss a lot, the same way I might in my own milieu.
Jorge Riechmann, the Spanish poet who translates Char, says that you should know the exact way in which Char uses a certain word in order to translate. So "humidité" might have a particular connotation for Char that is particular to him. The language within the language, or idiolect of a poet, Riechmann calls it. A lot of French readers of Char don't probably have that level of intimacy with his language, so it goes deep than "native speaker" knowledge.
I think it is important to simply say: you have to know what you're talking about. These are just examples of that.
Where I am going with this is a talk about Celan I have to give in Spain in May. I guess I am going after that "translation as deep reading" theory that I found implicit (and explicit) in the book of translation essays I am reading. I have to start my talk by saying I'm not a Germanist, but that Celan is a poet that can only be approached from a Comparative Literature perspective. I might look a bit at his Rumanian poems.
The deep knowledge goes both ways. You have to know the literary language & traditions into which you're translating as well as the literary language and traditions of the source. Just like, to write my books on Lorca that deal with American culture, I have to know both Lorca and American culture.