…which is the problem I was referring to: by optimising for that—your—use case, those of other people will invariably suffer.

We only have a single text field as the input; how are we supposed to guess whether you want to find an exact match of the phrase, a fuzzy match, at least one of the words provided, or any other possible variation? Also, are you interested in the content, the subject, the recipient, the sender address you used, a header field, an attachment, what have you? Do you want them ranked by the frequency of the word, or the position from the start of the text? Does it count those occurrences in quoted passages of previous mails downthread multiple times? What if it’s a stop word?

There are of course sensible ranking solutions and heuristics for these questions. I just want to highlight it’s not as trivial as it first sounds. Most mail clients probably don’t ship with a Lucene index—while they should.