A lot of people complain loudly and frequently about what they call "interpretive translation". Most of these people are criticizing what is commonly called dynamic translation, which includes translations that tend to translate the sense of an expression as opposed to favoring the formal properties of the sentence. Either favoring can obscure the other, and good translators will know how to find the right balance to express the original meaning best. But the complainers don't understand the complexities of translation very well, or they would realize that sometimes capturing the sense of the original means sacrificing the ability to capture its form. Thus translations such as the NIV, TNIV, or NLT will come under their wrath, and they will favor translations such as the ESV, NASB, or NKJV.
One deep irony of this is that lots of interpretive translation goes on in those translations that people are, following the ESV translators, now calling "essentially literal". Wayne Leman points out one example. A lot of these translations that are supposed to avoid interpretive translation do exactly that all the time in ways that their supporters consider the right way to translate those passages. Wayne's example is in capitalizing words like 'son' or 'man' when the translators interpret them to be referring messianically to Jesus. But this is indeed an interpretation, even if the interpretation is based on other scriptural passages that quote it and apply it to Jesus.
Someone wanting to translate this way might defend it on the grounds that interpretations based on other parts of the Bible are infallible and thus can serve as the good kind of interpretation. After all, if Hebrews or Acts quotes Psalm 2 about Jesus, then can't we be 100% sure that it's simply talking about Jesus? If we believe the Bible to be infallible in its quotations, then this kind of interpretation is God's own interpretation, and thus it's true. Whats right about this is that someone who takes the Bible to be infallible should see the Bible's quotation of itself as infallible, i.e. it couldn't be an error in quotation. What's wrong about this argument, however, is that our interpretation of what the quotation is doing might be wrong. If Acts 13 applies Psalm 2 to Jesus, that doesn't mean its original referent is Jesus. It might be referring to the Davidic line in general in most of what it says, with Jesus representing the ideal Davidic king and thus fitting into its reference but not encompassing the entirety of its reference. Those who capitalize the pronouns about Jesus or the word 'son' are thus engaging in the bad kind of interpretive translation in this case, because it might actually give the wrong result.
I find it deeply ironic that those who oppose dynamic translations will adamantly support this kind of ripping of a passage from its original context simply because they don't understand the nature of NT quotation of the OT. This is a far more interpretive translation with much more significant implications than it is to translate the sense of an expression in ordinary English when its original imagery sounds foreign to English speakers but whose sense captures exactly what the original was really saying. It's against the very spirit of essentially literal translations to add in a capitalization that contains great meaning when the original had no such capitalization. Yet the so-called essentially literal translations do this all the time, and their supporters seem completely unaware of the inconsistency here.
In the end, I recommend never capitalizing pronouns for God, and I'm not even sure we should capitalize 'son' except when it clearly and unambiguously stands as a title for Jesus' role in the Trinity, as Jesus uses it regularly in John. We should capitalize proper names, of course, and that includes the word 'God' when used as a name for God, and we should capitalize words at the beginnings of sentences of course (I say "of course", but in the world of Instant Messenger even that's starting to go in some contexts). There are other ways of translating interpretively that essentially literal translations engage in all the time that this won't solve, but this is an easy fix for this particular problem without sacrificing the general principles of essentially literal translation.