The question “top or bottom?” can be answered with “versatile,” to mean “either, depending on circumstances.” Another possible answer is “interleaved.” Here’s an argument that the proper answer should be “both” to maximize satisfaction.
Quoting from Wikipedia:
The main options [when replying to email] are interleaved posting, also called inline replying, in which the different parts of the reply follow the relevant parts of the original post, bottom-posting, in which the reply follows the quote, or top-posting, in which the reply precedes the quoted original message.
As to which is better, the following bit of Internet lore summarizes the consensus among educated audiences:
A: Because it messes up the order in which people normally read text. Q: Why is top-posting such a bad thing? A: Top-posting. Q: What is the most annoying thing in e-mail?
Meanwhile, uneducated audiences often live along with the standard behavior of their e-mail client program of choice, which often uses top-posting by default.
Until now, RFC 1855 provided a guideline towards a middle road:
If you are sending a reply to a message or a posting be sure you summarize the original at the top of the message, or include just enough text of the original to give a context. This will make sure readers understand when they start to read your response. Since [...] it is possible to see a response to a message before seeing the original[, g]iving context helps everyone.
In other words, the proposed strategy is to top first, succinctly and efficiently, then interleave for details.
And today I realized just before sleep another advantage to a brief top, not envisioned by the authors of RFC 1855: most modern e-mail clients, especially those on smartphones, have a default message display called “short view” which shows the first line of text from each e-mail next to the subject line. By topping succinctly before a more extensive interleaved answer, the essence of the message can be delivered at the short view.
And this post is not shamelessly attracting audiences by reusing keywords: I believe that the other answer to “top or bottom?” should be the same, and for similar reasons. I have discovered a truly marvelous proof of this, which this margin is too narrow to contain.