The short version is this: it's a hard read, but good. It's full of things to start thinking about, but of strange detours as well. Plivorte:
There was an extremely frustrating run, around that point in my reading, of almost pure enumeration of translational subtleties, very much like the bulk of Eco's "Experiences in Translation". Frustrating, because (a) I don't know the language any of his source texts are in (or sometimes, the destination text, when he's discussing a translation out of english, and sometimes both texts, in the case of several translations between french and german that he mentions) and (b) I can't glean any sense, any systematic meaning from the mass of assertions and criticisms and recommendations. I know that he's roundly convinced that to make a good translation, one must choose the right words. Well, what's the right word? Of course, he tells me, it's just the right one, le bon mot, le meilleur mot, the one that just so deftly captures the same meaning as the one in the source. The just-right phrase and rhythm and stylistic brush-stroke are those that do the same thing somehow, as if consummating some sort of magic trans-cultural kinship of voiced breath.
But I come away from all five hundred pages of it without feeling like I've learned quite as much as his introductory pedagogical ambition had me thirsting for. Yes, if you grow up, like Steiner did, surrounded by a world that is hurled into your eyes and into your ears through English, French, and German all at once, then you certainly will have an inkling of these three worlds' homology: you will have some personal sense of each one's habits of onomatopoiea, each one's music. So he is uniquely qualified to perform a certain species of translation; he uniquely knows what tremors in his soul this sheaf of English words elicits, and what he hears ringing just beyond the concretely audible in that passage in German.
So what? If we're to grant that translation is possible at all, that it is meaningful at all, we have to look beyond the merely idiolectic. Translation into English is translation into the many literacies of the readers of English, into the community. So why even pause to note the idiosyncratic associations of one polyglot's poesy? It's the stuff of memoirs, not of theories.
Despite my complaints, there's a couple of core theses that kept showing up through the book that I feel quite a bit of sympathy with, even if they're not perhaps so groundbreaking. One is the importance of paying attention to the fact that the extreme reductionism of the case of idiolects, at least as a counterexample, a corner case. Whatever theory of language you have, it must be comfortable with the fact that, in some basic sense, I face a merely quantitatively easier task when I talk to my neighbor rather than a frenchman. The other theme (which now that I think of it isn't really so different from the one just mentioned) is that translation is far more ubiquitous than we conventionally suppose in the process of mere thought.
Good stuff, anyway. Much better than some of his later essays I've read, which get awfully stridently curmudgeonly and with the "oh, kids these days" and all.