- 1. a poetic translation may also fail to be that nice but since it pretends to be poetic a reader might think that the original text is bad poetry (the argument is K.R. Norman's),
- 2. a literal translation may help people understand the text, even if the translator has not been able to completely understand it.
In some cases, a translation cannot be too literal. Translating a compound with a compound (for instance, yāgasya iṣṭasādhanatva and "desired-[end]-meansness of the sacrifice"), to name the extreme case, would be absurd and would not help any reader. Breaking a compound into a nominal phrase ("being the means for the desired end of the sacrifice") does not always work either. In such cases, one just has to acknowledge the fact that English is more a verbal language than Sanskrit and translate "the sacrifice is the means to realise a desired end".
That being said, the kind of translation you are writing depends on your target reader. If you write for a student or for someone who uses your translation just as a commentary to the Sanskrit text, you might want to enable him to check back any single word you are translating in the original text. If, on the other hand, you are writing for a reader who does NOT know any Sanskrit (or Pāli or any other language), then you should be able to produce a text which is independently understandable. An excellent example of the first kind of translation is Jim Benson's translation of the Mīmāṃsānyāyaprakāśa; of the latter John Taber's translation of the Ślokavārttika chapter on perception.
Is there space left for a third possibility?
I wrote by far too many posts on translations in the last months. You can find them listed here.