You have to look at the etymologies, not the spelling. Attach has a Latin ad- prefix, and the /d/ has become a /t/, which is still there in the spelling, but not in the pronunciation, like most English spelling. English spelling, in case you hadn't noticed, sucks. This is because it was invented 500 years ago and we feel that it should never be changed, even when it doesn't work. Kind of like Windows.

Make that an answer and you got some points. And wait, aren't they changing Windows?

@screenmutt Note the similar patterns in adduce/deduce, allude/delude, accelerate/decelerate, affect/defect, aggradation/degradation, append/depend, etc. The original d of the preverb ad- assimilates into many following consonants, but the preverb de- doesn’t have a final consonant that can assimilate to anything (it originally had a final s, but that was lost in early Old French before most consonants, and the preverb was levelled to just dé- everywhere).

@JohnLawler you mentioned in your first comment that attach is derived from the Latin's prefix ad-. Well according to three other posters this is false. Attach is derived from the French attachier. I find it boggling that users continue to upvote your comment even after the two answers posted, and mac's (now deleted) answer refuting this fact. Could you please confirm whether or not you stand by your statement. (I think in part it influenced Janus's well-written answer, which I still support and think he should undelete. And not because it is the "right" answer.)

It was actually more complex than what I said; every word has its own history, with its own arbitrary choices, just like every individual gene does. Several additional steps were involved, as @user77834 points out. The point, in any case, is that Latin morphology is not modern English morphology, and should not be expected to be. Just as Middle English spelling does not represent modern English pronunciation, and should not be expected to. Everybody would like grammar to be regular and predictable, just as everybody would like pi to be equal to 3. But it doesn't work that way.

@John Mack - Isn't this question based on a misconcept? Why should the two terms have the same number of Ts?

@Josh61. Precisely. Based on a misconcept? I don't know. Hovever, everyone's hesitation to lift their heads above the parapet here is the major reason for my interest in this question. That and the desire to find something to throw away my reputation points on (I'm averse to scores generally, although I've seen the cogent reasons for them in discussion here). Plus you've gifted me 'misconcept'. A gem, I'll use it frequently from now on.

@John Mack - are you aware of all the 'high standard' answers that were posted and then for some reason deleted under this question?

@Josh61 Well yes I read about that happening, but I never saw the content. I am under no illusion that a mere 400 points will lure them back out. As I understand it they were withdrawn as a matter of principle. Whatever it was I respect that.

I always thought the doubled "t" was intended to make the first "a" in "attach" short. The "e" in "detach" is supposed to be long, so it doesn't need an extra "t". But judging from the sizable debate here, I'm obviously way out of my depth.

After reading through the comments on the question, the various deleted answers, and the (deleted) comments on the deleted answers, I find it very amusing to see a bounty posted because "the question has not received enough attention".

Perhaps because the etymology of detach is a contraction of des and attachier the last letter of the first part and first syllable of second part were lost in the ordinary contraction of any word

@Hellion: And you might have added "the multiple bounties awarded to deleted answers." In fact, if I recall correctly, at least one additional bounty on a now-deleted answer was scrubbed (perhaps by a moderator) almost immediately after it was awarded, and so doesn't appear next to that answer. All in all, this question seems to have become the La Brea Tar Pit of EL&U.

@SvenYargs, just for the record, no bounty was 'scrubbed', anyone can see in the revisionsthat a bounty was awarded to a correct (but simple) answer but the poster turned it to a new reply that established the truth in an irrefutable manner, that triggered the withdrawal of the wrong reply.

@user119052 Please don't imagine I meant to impugn the character or motives of anyone involved at any time with this question. My choice of words ('matter of principle') was simply meant to suggest that the withdrawal of some material previously posted here was likely done not on a whim, but rather for good and reasonable, considered and honourable reasons Essentially this is a presumption, but one based on my high regard for my fellow posters, and specifically upon some of the positive comments previously posted here.

@JohnMack, the behaviour of some fellow posters was not exactly reasonable, considered and honourable: you cannot get the right picture since the clique who downvoted the correct answers and upvoted the wrong one (in good or ill faith) exposing themselves to ridicule quickly deleted all evidence. I just wanted to put the record right since you can't read the whole, real story: btw, the answer you rewarded quotes the same obsolete (disowned-on-printed-editions) entry of the OED that started it all

@user119052. Well I have a printed copy of the OED, so when I have some time on my hands (in a couple of days) I will check all the references. You've reminded me that it is good practice to do so. I can't say that the question or answer is central to my particular interest, but then again, exploring an obscure subject often opens up one's mind to new interests. A bounty, or indeed any reputation awarded, is no indication of 'correctness' or authority to my mind. I 'burn' my reputation deliberately by awarding bounties - recklessly perhaps, but it serves my purpose.