But automatic PGP/GPG encryption is perfectly plausible IF the user enables some scripts.
Even if that were technically true, it's irrelevant because nobody in their right mind would enable scripting on a site like this, especially in view of what happened at Freedom Hosting.
But just out of academic interest, could you explain what you meant by this:
There would be an algorithm sent to the client that generates a key-pair from a user-entered word/phrase. The user would enter the word/phrase to generate the private key on the client, and decrypt the message client-side.
... because it seems like you might be making one of the following mistakes:
(a) confusing symmetric with asymmetric encryption, or
(b) using the words:
"generate a key-pair from a user-entered word/phrase"
to mean:
"retrieve an existing private key from vendor's locally-stored keyring, given the passphrase for said key", or
(c) talking utter nonsense.
In your scenario, the buyer has encrypted his address with the vendor's public key, yes? So then, when the vendor wants to decrypt it, you're saying he generates a new key-pair which, by virtue of being derived from an appropriate user-entered word or phrase, happens to be identical to the vendor's own key-pair? That can't be right. How long would the phrase have to be, in order for that to work? (Answer: very long indeed, tantamount to requiring the vendor to memorize his entire private key, and to type it in every time he processes an order). This scenario would indeed be utter nonsense, that's partly what I meant by (c) above, and I just wanted to eliminate it first.
If the mistake you've made is (b), then fine I understand what you mean.
If it's (a), then I guess you're saying there are no actual PGP keys involved in the encryption or decryption, and that both buyer and vendor must enter the same word/phrase as each other (a pre-shared secret) and that the script would derive a *symmetric* key from it and use that key for encryption and decryption. Again this doesn't seem right; asymmetric encryption thankfully eliminates the need for pre-shared secrets, so why would we take such a backward step?
These are the only ways I have of interpreting what you said, and none of them seem quite right. So if it's something else, please explain.
Or don't, as you wish. I was just curious.
[Admittedly when I wrote the original comment I neglected to think of the specifics of implementing it (i.e. javascript would almost certainly be required unless additional plugins/scripts were installed), but I was certainly not suggesting a server-side approach.]
Ah, ok. (hastily editing for your edit, I didn't see that before).
From your last remark I see that we are probably in agreement, that this is more of an academic discussion, rather than something practical that a site like this might actually implement. Good then, and thanks for your reply.
