I'm familiar with four translations of the Mahabharata. The first of these, that of C. Rajagopalachari (1951), contains a selection of over one hundred of the more interesting stories and episodes with interspersed comment.
Rajagopalachari's translation is a joy to read. The quality of his English style is on a par with that of the finest English writers, and his vivid and dynamic versions of these stories capture much of their humor and poignancy, and have a real vigor, sparkle, and human interest. Anyone approaching the Mahabharata for the first time could do worse than start with this wonderfully readable selection.
I wish I could say the same for the second and longer abridgement that came my way, that of Kamala Subramaniam (1965) in 766 large pages. Sadly, though one appreciates the effort that went into it, this is a book that I could not in good conscience recommend to anyone. Subramaniam seems to have had no grasp of English style at all. She has chopped each Sanskrit verse into small bite-sized pieces of English, and the staccato effect of her unending series of short, simple, unvaried sentences would, I think, weary any discerning reader.
The third translation, and the only complete one I have, is that by Kisari Mohan Ganguli, published between 1883-1896. Mine is the economy paperback reprint in four stitched and sturdily-bound though poorly printed volumes, and runs to over 5000 closely printed pages. Although not, of course, based on the recent critical Poona edition of the Sanskrit text, this edition should serve well enough as a reading text for anyone but a Sanskrit scholar.
J.A.B. van Buitenen, in the first volume of his own recent translation, comes down rather hard on Ganguli, though he apologizes for his harshness in a later volume. But to an impartial reader, van Buitenen's harshness seems hardly justified. As a native speaker of English myself, I find Ganguli's feeling for English to be on the whole superior to that of van Buitenen.
We should also remember that Ganguli did not have access to the rich resources van Buitenen enjoyed. In addition, Ganguli states clearly in his preface that he has tried to give "as literal a rendering as possible of the great work of Vyasa," and a literal rendering does not have the same aim as a more literary rendering.
The most prominent feature of Ganguli's style, apart from its literalness, is his employment of forms such as "Thee" and "Thou" and "Thine," etc., archaic forms which can at times grate on the modern sensibility.
Despite his literalness and archaisms, however, and despite his occasional inaccuracies (some of which seem to be the product of misprints), Ganguli is always lively and never wooden; as an Indian, he seems really to have caught the spirit of the Mahabharata. His version, though it requires stamina to read, has great energy and succeeds marvelously in capturing the many interesting and colorful characters of the poem, and in vividly portraying the weird and wonderful things they get up to. Ganguli's is a lively edition I would certainly recommend.
As for the more recent three volumes of van Buitenen's translation (1973-78), which cover just one third of the total text (Books 1 to 5 of 18), although they represent fairly careful and up-to-date scholarship, and although they are beautiful examples of a well-thought-out layout and typography which makes for much easier reading than the cluttered pages of Ganguli, stylistically they too leave something to be desired, at least occasionally. Van Buitenen had his quirks too.
His grasp of the connotations of English words is often weak, and sometimes I even get the feeling that he may not have been a native speaker of English. Why else such eccentric usages as "Prince sans blame," or "The Age of the Trey" and "The Age of the Deuce"? Even worse, why "Baron," with its wholly inappropriate medieval European connotations, instead of the Sanskrit "ksatriya" or the English "Warrior"? A European "Baron" suggests to me something very unlike an Indian "ksatriya." Far better to keep occasionally to the Sanskrit vocabulary, which is simple enough, than flee to inappropriate equivalents.
Besides van Buitenen's occasionally quirky usage, it must be said that his rendering can sometimes be rather wooden, particularly in the passages he chose to attempt in 'verse.' On the whole, however, he has given us a version which at its best reads well, and one that is mercifully free of irksome archaic forms. His edition is also extremely well-organized, and has a substantial and helpful scholarly apparatus (lengthy introductions, plot summaries, notes, full indexes, etc.) which Ganguli's edition lacks.
So where are we? Clearly no ideal and complete English translation of the Mahabharata exists, nor is ever likely to exist given its stupendous size. Also, to really get a feeling for the magic of the Mahabharata, you have to read at least a bit of it in Sanskrit. A practical and user-friendly 'Introduction to Sanskrit' for ordinary folks (as opposed to academic linguists) is that of Thomas Egenes (1989). A few months work with this will soon find anyone reading at least some of the Sanskrit, in a bilingual edition such as Monier Williams' excellent 'Story of Nala,' with real enjoyment.
To conclude, if I had to choose between the Ganguli and van Buitenen, and although I'm grateful for both as both have much to offer, I would recommend Ganguli as being closer in spirit to the original - but I'd also suggest that those who are innocent of Sanskrit take a peek at Egenes Introduction to Sanskrit, Part 1.