I read a couple of books on translation this summer: Le Ton beau de Marot by Douglas Hofstadter and Dire presque la même chose by Umberto Eco. Although both book titles are French, Hofstadter's work is in English, and I read a French translation (from the Italian) of Eco's work.
Le Ton beau de Marot is a multitude (80+) translations of the same 28-line poem by Clément Marot, interleaved with musings on translation and meaning. Hofstadter's translation problem can be summed up as "What happens where you take something from over there and put it over here". What does something mean if meaning is tied to context and the context changes? How can we transmit meaning without transmitting the entire context? Where does this do to the problem of artificial intelligence and computer understanding? No translation is exact -- some aspect of the source text is lost because of the differences between the source context and the destination context. The job of the translator is to find the best mapping between the two, and to choose which aspects are important and which can be discarded or minimized. The eighty or so translations come to play as examples of choosing to honour (or not) different aspects of Marot's original poem.
I loved Le Ton beau de Marot. I think this book really benefits from the reader having an understanding of French, although it's not strictly required. Hofstadter speaks French fluently and his love of French comes across in the book. As a language learner, I also really enjoyed reading about his experiences learning and speaking foreign languages. Finally, the chapter discussing the process of translating Gödel, Escher, Bach into Chinese and French made me more excited to read the French version. The challenges here were also interesting, since some of GEB is autobiographical. Hofstadter had to negotiate with his translators to "forge" certain aspects of his life to capture the larger ideas he wanted to get across, while in other cases details of Americana and his life were given importance over a pun or other wordplay.
Eco's book, Dire presque la même chose has essentially two parts. The first half deals with his experiences translating and being translated; the second part is more academic and delves deeper into the theory of semiotics and how it relates to translation work. As Le Ton beau de Marot benefits from a knowledge of French, so I think would Dire presque la même chose benefit from a knowledge of (at least) Italian. Most of Eco's examples from the first half of the book are edge cases and shades of meaning either from or to Italian. I say "at least" Italian since there are plenty of example translations to and from German, English, French and Spanish as well.
I enjoyed Dire presque la même chose less. Eco is a very erudite writer, and so many of the examples from his works are similarly dense and obscure. I probably wouldn't have understood them even if they were in English. Discussions of shades of meaning between words in Italian and the word the Spanish translator chose is not that interesting for me as I don't speak either of those languages. The second half of the book I found a bit too beyond my level in terms of the semiotics discussed. Again, I'm not sure I would have enjoyed it had it been in English, although I would have been able to skim it faster.
One interesting thing I did get out of Dire presque la même chose was the fact that Eco wrote The Name of the Rose to be deliberately difficult to read. That makes me feel better about not understanding it.