[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Zimmermann/NSA debate postponed



Carl said:

> Isn't it time for cypherpunks-who-write-code to respond to this obvious
> customer need?
> 
> I have no mailer myself but I do have a friend who produces one and I'm
> helping him to incorporate PGP seamlessly.  It's taking a long time but it
> should be worth it.
> 
> Anyone else out there with their own mailer?

I've fooled around with writing a mailer, and I ran up against a few 
walls.  The biggest one was my not being the world's greatest programmer, 
and not knowing much about email systems.  But there are other problems 
as well, that I think would plague anyone who tries to write a mailer.

The goal that I was working for was pretty straight forward:  some sort 
of a secure system that would be as transparent as possible.  Ideally, it 
would work just like elm or pine or eudora:  you'd just say "I want to 
send mail to so-and-so", and all the rest would happen automatically.  
The mail would be encrypted, your signature would be affixed, the 
recipient wouldn't have to worry about decrypting the mail, and there'd 
be some small little indicator on his status bar that said the signature 
was good.

The first thing I realized when I started fooling around with this is 
that my basic design, which was modeled on a QWK packet offline reader, 
wasn't practical.  Why?

First of all, all of the crypto work has to be done on a machine
controlled by the user.  This is obvious, and it's the reason I wanted to
go QWK style in the first place.

But think about what happens if you dl a packet of incoming mail.  
Inside, you've got a letter from someone you've never met before, and 
it's signed.  How can you verify the signature?

Right now, people don't use secure mail for day to day traffic because
it's too much of a hassle.  You can write a mailer to automate encryption
and decryption, affixing and verifying signatures.  But you're still going
to have to require users to hunt down keys, decide if they want to trust
them, and load them onto their key rings.  That's enough hassle for most 
people to stick with elm.

It seems to me that a prerequisite for a transparent, secure mail system
is an efficient, interactive, IP based key distribution system.  It would
have to be distributed, at least if we were serious, because it would have
to be able to handle several million people's keys.  And the web of trust 
would have to be such that keys could be accepted (or rejected) 
automatically, without human intervention.

If such a system of keyservers existed, it's not hard to imagine really
useful secure mail systems, although implementing them would still be an
awful lot of work.  Sendmail could be configured to feed mail through
programs to handle the crypto work on single user unix machines (or even
on multi-user machines, if the users trusted the sysadmin), and some sort
of a secure pop system could be developed for other people.

But this would be a very big project, and would necessitate agreements on 
standards, etc.