All you would need for this is to model each user's interaction with the system as a statement, where a statement could be either code or data, and then sign and encrypt these statements by default and encrypt the statement's decryption key with recipients' public keys as needed.
This might sound esoteric but we could already do this today just by taking something like Unison [1] and adding this encryption layer.
As geeks we should really own up to the fact that it's perfectly feasible to do this... and get working on it.
P.S. For a fun fact; at it's core such a language would pretty much be what John McCarthy described as 'the language for the year 2015.' [2]
[1] http://unisonweb.org [2] http://lambda-the-ultimate.org/node/3180
It's integrated with Thunderbird via Enigmail. It works at least as well as MS Outlook's integrated message signing. If your reply to that is something like "Everyone uses webmail.", I say "I've been using IMAP to access Gmail since five minutes after the feature was added to Gmail.".
This[0] seems to indicate that the only email client that has real problems with attached signatures is Outlook Express. I can't bring myself to care much about Outlook Express users. (Is OE even installed on Windows 7 and later?)
The bigger problems are that it’s not integrated and it’s not effective for shielding metadata. With difficulty and being mindful of the sharp edges, you can use OpenPGP on desktop. You need a clunky third-party email client to get it on mobile. Best case scenario, you send lots of extra data that gets ignored. Worst case, all those attached signatures get downloaded as files, causing confusion and anger. And there’s no interoperability with S/MIME, which actually is supported by most proprietary email clients.
OpenPGP does not protect metadata. It piggybacks on SMTP, which publishes the sender, recipient, and subject line in the clear, along with the identity and timing of each server in the transport path from when you send the email to when it lands in your recipient’s mailbox. PGP was originally conceived as the “envelope” to protect your message contents, but in this era of unlimited surveillance we need to do better.
"Why aren't we doing this now?" Probably for the same reasons that we're going back to the 1990's world of Instant Messaging walled gardens; techies and cypherpunks are winning some battles but not most of them, and non-technical users don't understand what they give up when they choose a centralized Web and ISPs that make them incapable of acting as a peer on The Internet. [0]
To the rest of your comment:
To the best of my knowledge, I've never had Enigmail or Thunderbird mangle my PGP signatures. I participate in a few technical mailing lists; the folks there absolutely would tell me if my signatures were getting fucked up. For mail that doesn't go to these lists, I can double-check the message copied to my sent mail folder. :)
Installation of Enigmail on Windows, Linux and -I presume- Mac is trivial. Based on reports, the sharp edges that you speak of appear to be there; send only plaintext email and attach -rather than inline- your signature and you will -in my experience- avoid all of them.
The world of mobile software is largely a cesspool. Maybe I'm ignorant, but it seems like the only folks doing good privacy protection mobile software are Open Whisper Systems, the Tor Foundation (by way of Orbot), and Whatsapp. (The Whatsapp folks are in this list because of the work that Open Whisper Systems did to integrate TextSecure's near-zero-effort crypto into Whatsapp's software.)
[0] Yes, I totally understand that almost no one in the US has a real choice in ISP. Even here in Silicon Valley, I find myself prevented at every turn from giving my money to local, independent ISPs. There's a little comfort in the fact that Comcast allocates and routes up to a /60, and performs very, very little inbound filtering.
which fatal flaws are present in "IPsec"?
No, the answer includes better tech. OpenSSH and LibreSSL are doing good work in removing insecure codecs. HTTP/2 looks like it should do well, because no client implementation accepts plaintext. I’m hoping that protocols like OTR and Darkmail become more widespread.
PGP is basically unusable. You can use it to dump secrets on journalists, after you have trained the journalist. You can’t use it to lead a quiet, private life. I tried. It’s not integrated with anything, and nobody else uses it. It’s fatally flawed.
Being riddled with fatal flaws is an aspect PGP shares with the lucrative Certificate Authority system, with S/MIME, with the DNS security system, with IPsec, with cell phone encryption.
Worse, actual solutions are orthogonal to big business interests. Decentralized trust? Google has no need for this; Google has its own intermediate CA (which, through the totally broken CA model, means Google can issue valid TLS and S/MIME certificated for any domain in the world). Federated content sharing? Facebook has no need for this; Facebook needs all your data to keep it safe for you, and also to analyze and monetize it.
What we need is to follow the article’s plea: Truly understand that the network is hostile, and design our systems to make security the easy default. And drop insecure protocols already, wherever there is a viable alternative.