Copyright (C) 1995 Patrick J. LoPresti
The Mailcrypt program and this documentation are published as free software. You may redistribute and/or modify them under the terms of the GNU General Public License as published by the Free Software Foundation; either version 2, or (at your option) any later version.
Mailcrypt is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details.
You should have received a copy of the GNU General Public License along with GNU Emacs; see the file COPYING. If not, write to the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA.
Mailcrypt is an Emacs Lisp package which provides a simple but powerful interface to cryptographic functions for mail and news. With Mailcrypt, encryption becomes a seamlessly integrated part of your mail and news handling environment.
This manual is long because it is complete. All of the information you need to get started is contained in this Introduction alone.
Mailcrypt requires version 19 of GNU Emacs. Mailcrypt has been tested on a variety of systems under both FSF Emacs and XEmacs.
Mailcrypt requires Pretty Good (tm) Privacy, usually known as PGP. This document assumes that you have already obtained and installed PGP and that you are familiar with its basic functions. The best way to become familiar with these functions is to read the PGP User's Guide, at least Volume I.
For more information on obtaining and installing PGP, refer to the MIT PGP home page at http://web.mit.edu/network/pgp.html.
Although Mailcrypt may be used to process data in arbitrary Emacs buffers, it is most useful in conjunction with other Emacs packages for handling mail and news. Mailcrypt has specialized support for Rmail (see section `Reading Mail with Rmail' in The GNU Emacs Manual), VM (see section `Introduction' in The VM User's Manual), MH-E, and Gnus (see section `Overview' in The Gnus Manual). Information on the general use of these packages is beyond the scope of this manual.
If Mailcrypt is not installed on your system, obtain the latest version from the Mailcrypt home page at http://cag-www.lcs.mit.edu/mailcrypt/ and follow the instructions in the file `INSTALL'.
Next, teach your Emacs how and when to load the Mailcrypt functions and install the Mailcrypt key bindings. Almost all Emacs major modes (including mail and news handling modes) have corresponding "hook" variables which hold functions to be run when the mode is entered. All you have to do is add the Mailcrypt installer functions to the appropriate hooks; then the installer functions will add the Mailcrypt key bindings when the respective mode is entered.
Specifically, begin by placing the following lines into your `.emacs' file (or the system-wide `default.el' file):
(autoload 'mc-install-write-mode "mailcrypt" nil t) (autoload 'mc-install-read-mode "mailcrypt" nil t) (add-hook 'mail-mode-hook 'mc-install-write-mode)
Then add additional lines for your own mail and news packages as described below.
To hook Mailcrypt into Rmail, use the following lines:
(add-hook 'rmail-mode-hook 'mc-install-read-mode) (add-hook 'rmail-summary-mode-hook 'mc-install-read-mode)
To hook Mailcrypt into VM, use the following lines:
(add-hook 'vm-mode-hook 'mc-install-read-mode) (add-hook 'vm-summary-mode-hook 'mc-install-read-mode) (add-hook 'vm-virtual-mode-hook 'mc-install-read-mode) (add-hook 'vm-mail-mode-hook 'mc-install-write-mode)
To hook Mailcrypt into MH-E, use the following lines:
(add-hook 'mh-folder-mode-hook 'mc-install-read-mode) (add-hook 'mh-letter-mode-hook 'mc-install-write-mode)
To hook Mailcrypt into Gnus, use the following lines:
(add-hook 'gnus-summary-mode-hook 'mc-install-read-mode) (add-hook 'news-reply-mode-hook 'mc-install-write-mode)
All Mailcrypt commands are (by default) activated by three-character key sequences which begin with C-c /. The four most common operations are:
These functions and others are documented in detail in the following chapters.
Any time you are composing or reading mail or news, you can get a
summary of the available commands by typing C-h m. If you are
running Emacs under X, an even easier way to see the available commands
is to access the Mailcrypt
pull-down menu.
Mailcrypt works by providing two minor modes for interfacing with
cryptographic functions: mc-read-mode
and mc-write-mode
.
mc-read-mode
provides key bindings for processing messages which
you have received; mc-write-mode
provides key bindings for
processing messages which you are about to send. These minor modes will
indicate when they are active by placing a characteristic string in the
mode line (see section Mode Line). They will also add a Mailcrypt
pull-down menu to the menu bar.
The normal installation procedure (see section Installation) will arrange
for the appropriate mode to be active when you read and compose mail and
news. But you may want to use Mailcrypt's functions at other times; to
do so, you can call mc-install-read-mode
or
mc-install-write-mode
directly. For example, if you were editing
a file in Text mode and wanted to digitally sign it, you would type
M-x mc-install-write-mode, then C-c / s (see section Signing a Message).
Once one of the Mailcrypt modes is active, you can get a summary of the
available functions by typing C-h m or by examining the
Mailcrypt
pull-down menu.
The description of each function below includes which of the modes has a binding for that function.
The function mc-encrypt
will encrypt a message in the current
buffer. mc-write-mode
binds this function to C-c / e by
default.
When this function is called, Mailcrypt will prompt you for a comma-separated list of recipients. If called from a mail composition buffer, the recipient list will default to the Email addresses in the `To', `CC', and `BCC' lines of the message.
If you want to be able to decrypt the message yourself, you need to add
yourself to the recipient list. If you always want to do so, set the
variable mc-encrypt-for-me
to t
. (Note that Mailcrypt
overrides the PGP "encrypttoself" flag; use this variable instead.)
If you provide an empty recipient list, Mailcrypt will ASCII-armor the message without encrypting it.
Once you have edited the recipient list to your satisfaction, type
RET to accept it. You will then be asked whether you want
to sign the message; answer y or n. You can avoid this
question by setting the variable mc-pgp-always-sign
: A value of
t
means "yes", a value of 'never
means "no".
If you elect to sign the message, Mailcrypt will prompt you for the appropriate passphrase unless it is cached (see section Passphrase Cache).
Mailcrypt will then pass the message to PGP for processing. Mailcrypt
will call the functions listed in mc-pre-encryption-hook
and
mc-post-encryption-hook
immediately before and after processing,
respectively. The encrypted message will then replace the original
message in the buffer. You can undo the encryption with the normal
Emacs undo command C-x u (see section `Undoing Changes' in The GNU Emacs Manual).
If an error occurs, Mailcrypt will display an appropriate diagnostic. If you do not have the public key for one of the specified recipients, Mailcrypt will offer to try to fetch it for you (see section Key Fetching).
The default key for signing is the first one on the secret key ring
which matches the string mc-pgp-user-id
; this defaults to
(user-login-name)
. Note that this differs from PGP's normal
default, which is to use the first of all of the secret keys. To
mimic PGP's behavior, set this variable to ""
.
If you want to use a secret key other than your default for signing the
message, pass a prefix argument to mc-encrypt
. (That is, type
C-u C-c / e.) Mailcrypt will prompt for a string and will sign with
the first key on your secret keyring which matches that string. It will
be assumed that you want to sign the message, so you will not be
prompted.
The function mc-sign
will clearsign a message in the current
buffer. mc-write-mode
binds this function to C-c / s by
default.
When this function is called, Mailcrypt will prompt you for the appropriate passphrase unless it is cached (see section Passphrase Cache).
Mailcrypt will then pass the message to PGP for processing. Mailcrypt
will call the functions listed in mc-pre-signature-hook
and
mc-post-signature-hook
immediately before and after processing,
respectively. The signed message will replace the original message in
the buffer. Do not edit the message further with the signature
attached, because the signature would then be incorrect. If you
discover you need to edit a message after you have signed it, remove the
signature first with the normal Emacs undo command C-x u
(see section `Undoing Changes' in The GNU Emacs Manual).
The variable mc-pgp-user-id
controls which secret key is used for
signing; it is described in section Encrypting a Message.
To use a different secret key, pass a prefix argument to mc-sign
.
(That is, type C-u C-c / s.) Mailcrypt will prompt for a string
and will sign with the first key on your secret keyring which matches
that string.
The function mc-insert-public-key
will extract a key from your
public keyring and insert it into the current buffer.
mc-write-mode
binds this function to C-c / x by default.
This function is useful for sending your public key to someone else or
for uploading it to the key servers (see section Key Servers). The inserted
key will be the first one on your public key ring which matches the
string mc-pgp-user-id
(see section Encrypting a Message).
You may want to insert a different public key instead; for example, you
may have signed someone's key and want to send it back to them. To do
so, pass a prefix argument to mc-insert-public-key
. (That is,
type C-u C-c / x.) You will be prompted for a string; the first key
on your public key ring which matches that string will be inserted.
The function mc-decrypt
will decrypt a message in the current
buffer. mc-read-mode
binds this function to C-c / d by
default.
When this function is called, Mailcrypt will prompt you for the appropriate passphrase unless it is cached (see section Passphrase Cache).
The encrypted message will then be passed to PGP for processing. If you are not in a mail buffer, the decrypted message will replace the encrypted form. If you are in a mail buffer, you will be prompted whether to do the replacement.
If you answer n, you will be placed in a new mail reading buffer to view the decrypted message. This new mail reading buffer will have no corresponding disk file; its purpose is to provide you with all of your usual reply and citation functions without requiring you to save the message in decrypted form. Type q to kill this buffer.
You can avoid the question of whether to replace the encrypted message
by setting the variable mc-always-replace
. A value of t
means "yes"; a value of 'never
means "no".
If the encrypted message is also signed, PGP will attempt to verify the signature. If the verification fails because you lack the necessary public key, Mailcrypt will offer to fetch it for you (see section Key Fetching).
Look in the *MailCrypt*
buffer to see the result of the signature
verification.
The function mc-verify
will verify the cleartext signature on a
message in the current buffer. mc-read-mode
binds this function
to C-c / v by default.
When this function is called, Mailcrypt will pass the message to PGP for processing and report whether or not the signature verified.
If the signature failed to verify because you lack the necessary public key, Mailcrypt will offer to fetch it for you (see section Key Fetching).
The function mc-snarf
will add to your keyring any keys in the
current buffer. mc-read-mode
binds this function to C-c / a
by default.
This function is useful when someone sends you a public key in an Email message.
This is a long chapter describing an advanced feature; you may want to skip it on first reading.
There are several anonymous remailer services running on the Internet. These are programs that accept mail, strip off information that would identify the origin of the message, and forward the mail to the designated recipient. This simple scheme alone, however, is insecure if the anonymous remailer becomes compromised (or if the remailer was set up by an untrustworthy party in the first place). Whoever controls the remailer will have access to the identities of senders and recipients.
One solution to this is to use chains of remailers that send encrypted messages. For example, suppose Bill wishes to send a message to Louis using a chain of remailers A, B, and C. He writes the message (possibly encrypting it for Louis), then encrypts the result (including the fact that Louis is the recipient) using a public key supplied by remailer C. Then he encrypts this result using a public key supplied by remailer B. Then he encrypts this result using a public key supplied by A and sends the message to A.
When A receives the message, it decrypts the message with its key to produce something encrypted for B, learns that the next remailer in the chain is B, strips off the information that the message came from Bill, and sends the message on to B. B then decrypts, learns that the next remailer in the chain is C, strips off the information that the message came from A, and sends the result to C. C then decrypts, learns that the destination is Louis, strips off the information that the message came from B, and sends the result to Louis. With this arrangement, only A knows that the original message came from Bill, and only C knows that the intended recipient is Louis. In general, the sender and recipient can both be known only to someone who has compromised all remailers in the chain.
If Bill wishes, he can include an encrypted "response block" in his message to Louis, which defines a remailer chain that Louis can use to reply to Bill. Louis can use this chain without knowing who Bill is -- only the last remailer in the chain need know the final recipient. Bill can also establish a pseudonym for use in signing his anonymous messages.
Mailcrypt includes facilities for sending messages via remailers, for defining chains of remailers, for generating response blocks, and for using pseudonyms.
To use Mailcrypt's remailing facilities, you need to configure them first. Begin with the following steps:
finger
pgpkeys@kiwi.cs.berkeley.edu
for a list of remailer public keys. Note
that Mailcrypt requires that you have the public keys of all the
remailers you want to use, and therefore that the remailers support PGP
encryption.
Note: These steps need only be done once, although repeating them from time to time is probably a good idea, since remailers come and go.
Now test the remailer functions. First compose an outgoing Email message (using C-x m, for example) addressed to yourself. Type C-c / r. Choose a remailer; use TAB to get completion on its name. The buffer will be rewritten for anonymous mailing through that remailer.
mc-write-mode
binds the function
mc-remailer-encrypt-for-chain
to the key C-c / r. This
function rewrites the message for a remailer or chain. The resulting
buffer is just a new Email message, so it can itself be rewritten for
another remailer; this is one way to manually construct a remailer
chain.
Mailcrypt also has powerful facilities for defining automatic chains. We will start with an example. Suppose you have put the following into your `.emacs' file:
(setq mc-remailer-user-chains '(("Foo" "alumni" "robo") ("Bar" (shuffle-vector ["replay" "flame" "spook"])) ("Baz" "Foo" "Bar" "rahul" "Bar") ("Quux" 4)))
This code defines four chains. The first is named "Foo" and consists of "alumni" and "robo", in that order. The second is named "Bar" and consists of "replay", "flame", and "spook" in some random order (a different order will be chosen each time the chain is used). The third is named "Baz" and consists of 9 remailers: The two from "Foo", followed by a permutation of the three from "Bar", followed by "rahul", followed by another permutation of the three from "Bar". Finally, the fourth is named "Quux" and consists of a random permutation of the four best remailers as ordered in the `~/.remailers' file.
Now whenever you are prompted for a "remailer or chain", the chains "Foo", "Bar", "Baz", and "Quux" will be available, including TAB completion on their names. By capitalizing their names, you guarantee they will show up near the top of the completion list if you type TAB on an empty input.
Now for the gritty details. mc-remailer-user-chains
is a list of
chain definitions. A chain definition is a list whose first element is
the name (a string) and whose remaining elements form a remailer
list. Each element of a remailer list is one of the following:
So, in the example "Bar" above, shuffle-vector
is actually a Lisp
primitive which returns a random permutation of the argument vector.
(Which brings up a side note: A remailer list can be a vector instead of
a list if you like.)
So where do the definitions for "replay" etc. come from?
There is another variable, mc-remailer-internal-chains
, which has
the same format as mc-remailer-user-chains
. In fact, the
concatenation of the two is always used internally when resolving chains
by name. The "internal chains" are normally generated automatically
from a Levien-format remailer list, which lives in `~/.remailers'
by default and is parsed at startup time. The parser creates several
chains, each containing a single remailer, and names each chain after
the respective remailer.
Thus "replay" (for example) is actually the name of a chain whose single element is the remailer at <remailer@replay.com>. So "replay" is a valid name of a chain to include in the definition of another chain, as was done above in the definition of "Bar".
Mailcrypt can generate a response block for you. Just type C-c / b in an outgoing mail buffer. That will prompt you for a chain to use, and will insert the response block at point. Note that you can use any chain you want for your response block; it need not be related to the chain you (later) use to remail the message.
If instead you type C-u C-c / b, you will be dropped into a recursive edit of the innermost part of the response block. This text is what you will see at the top of the message when the response block is used. This text is the only way to identify the response block, since it will be used to mail you through anonymous remailers.
You probably won't need to use the C-u feature, since by default the response block contains the date, `To' field, and `From' field of the message you are composing. However, if you want your response block to point to a USENET newsgroup instead of your Email address, you may edit the innermost part of the response block to have a `Newsgroups' line instead of a `To' line.
Inserting a response block also updates the `Reply-to' hashmark header field. So, when your recipient replies to your message, the reply will automatically be addressed properly. This only works if the last remailer in the chain used to encrypt the message supports hashmarks (the response block chain doesn't matter). If the last remailer does not support hashmarks, Mailcrypt will generate an error when you try to use the chain.
Note that you should insert your response block before you encrypt the message for remailing. Also, see section Remailer Security.
Mailcrypt supports pseudonyms. Type C-c / p in an outgoing message buffer and you will be prompted for a pseudonym to use. Your pseudonym will show up in the `From' line that the recipient sees. Your pseudonym may either be a complete `From' line (including an Email address), or just a full name (with no Email address). In the latter case, the Email address will automatically be set to <x@x.x>, an invalid address designed to prevent sendmail from going rewrite-happy.
If you have one or more pseudonyms which you normally use, and you aren't afraid of revealing them if your account is compromised, you can set up a default list of pseudonyms with lines like the following in your `.emacs' file:
(setq mc-remailer-pseudonyms '("Elvis Presley" "Vanna White" "Charles Manson"))
Then those names will be available for completion when you are prompted for your pseudonym.
You should insert your pseudonym before you insert a response block, so that the response block will contain the `From' line as well as the `To' line. That way you can tell who you were pretending to be when you get a reply to your message.
Note: Many remailers do not support pseudonyms. In addition, the Levien format does not (yet) indicate which do and which do not, so Mailcrypt can't warn you when your pseudonym isn't going to work. The only way to be sure is to send yourself a test message, and to try different remailers until you find one or more which work. On the bright side, only the last remailer in the chain needs to provide such support; none of the others matter.
Mailcrypt knows how to rewrite USENET posts for anonymous or pseudonymous remailing. Just compose your post or followup normally, and use C-c / r to rewrite it for a remailer chain. You don't even need to start your newsreader to make a post; you can just compose a message in mail mode and replace the `To' line with a `Newsgroups' line before doing C-c / r.
Mailcrypt will generate an error if the last remailer in the chain does
not have both the post
and hash
(hashmarks) properties.
The hashmarks are used to preserve `References' and similar
headers, so your anonymous or pseudonymous followups will thread
properly. The variable mc-remailer-preserved-headers
controls
which headers are preserved when rewriting a message, but you should not
need to change it since the default value is reasonable.
Before rewriting, you can use C-c / p to insert your pseudonym, and C-c / b to insert your response block, just like when composing mail. In this case, the response block will include the `From' line and the `Newsgroups' line (which is the news analogue to the `To' line).
Mixmaster is a new kind of remailer which provides excellent security against traffic analysis and replay attacks. (For more information on these attacks and Mixmaster, see Lance Cottrell's home page at http://www.obscura.com/~loki/.
If you do not use Mixmaster, you may skip this section entirely; Mailcrypt's default configuration treats Mixmaster as if it did not exist.
If you have the Mixmaster executable installed, you can tell Mailcrypt to use it by placing lines like the following into your `.emacs' file:
(setq mc-mixmaster-path "mixmaster") (setq mc-mixmaster-list-path "/foo/bar/baz/type2.list")
mc-mixmaster-path
is a string representing the Mixmaster
executable. mc-mixmaster-list-path
is the complete path to the
type2.list
file.
Once these variables are defined, Mailcrypt will automatically try to
use the Mixmaster executable whenever possible. Specifically, when you
rewrite a message for a chain, Mailcrypt will find maximal length
sub-chains which have the mix
property and will use the Mixmaster
executable to rewrite for those sub-chains.
This allows arbitrary intermingling of Mixmaster and normal (also called Type 1) remailers, but you should note that this is not recommended. The recommended procedure is to have a single Mixmaster sub-chain which is most or all of the whole chain.
There are advantages and disadvantages to having the Mixmaster sub-chain at the end of the whole chain. The primary advantage is that Mixmaster remailers support multiple recipients. The primary disadvantages are that they do not support pseudonyms nor posting.
So here, as always, it is the last element of the chain which needs to support the special features you want. In general, the remaining elements do not matter, and the superior security of Mixmaster remailers is a good argument for using them for the bulk of your chains.
Mixmaster remailers also have a "Type 1 compatibility mode" which you
might want to invoke to use a pseudonym or make a post. You can do this
with the function mc-demix
. Here is an example of its use:
(setq mc-remailer-user-chains '(("Foo" "vishnu" "spook") ("Bar" "Foo" (mc-demix "replay"))))
This makes "Bar" a chain of three remailers, and guarantees that the last one ("replay") will be used in compatibility mode.
Note that Mixmaster remailers cannot be used for response blocks.
Mailcrypt will ignore the mix
property when generating a response
block.
Keep in mind that there is only one person fully qualified to protect your privacy: you. You are responsible for obtaining a list of remailers and their public keys; you are responsible for choosing which of them to use and in what order. There are public lists of remailers and keys (the Quick Start section above relies on them), but you pay for the convenience by putting your trust in a single source. This is one reason Mailcrypt does not access these public lists automatically; you need to get into the habit of watching what goes on behind the scenes. You should also try to learn something about the remailers themselves, since you are relying on them to help protect your privacy.
How many remailers should you include in your chain, and how should you choose them? That depends on whom you perceive as a threat. If the threat is your ex-spouse or your boss, even a single remailer is probably adequate (more won't hurt, but will cost in latency). If the threat is the Church of Scientology, you probably want to use a fair number of remailers across multiple continents. If the threat is a major world government, well, best of luck to you.
Also, there is a huge difference between chains suitable for regular messages and chains suitable for response blocks. Some remailers don't even keep mail logs (at least, their operators claim they do not), so it may be literally impossible to trace a message back to you after the fact if you chain it through enough remailers. Response blocks, on the other hand, have your identity buried in there somewhere. In principle, at least, it is possible to compromise the keys of all the remailers in the chain and decrypt the response block. So you should either use very long and strong chains for your response blocks, avoid using response blocks at all, or only use response blocks which themselves ultimately point to a newsgroup.
Here is a plausible sequence of operations when using the remailer support in Mailcrypt:
mc-sign
function
understands pseudonyms.
Now the recipient(s), reading your message through mail or news, can verify your pseudonymous signature; thus you have started to create a verifiable pseudonymous identity. If you use it consistently, it will develop a reputation of its own. With Mailcrypt, using a pseudonym is almost as easy as using your real name (and your followups in news will even thread properly). Welcome to the new age of letters...
This is a collection of tips for using Mailcrypt's remailer support.
mc-levien-file-name
for a description of Levien
format.
pgp
(required), hash
(required if you use hashmark headers), and post
(required for
posting to USENET). Remailers which do not support PGP won't even show
up in the completion list.
Mailcrypt can remember your passphrase so that you need not type it
repeatedly. It will also "forget" your passphrase if it has not been
used in a while, thus trading some security for some convenience. You
can tune this tradeoff with the variable mc-passwd-timeout
, which
is a duration in seconds from the last time the passphrase was used
until Mailcrypt will forget it. The default value is 60 seconds.
So, for example, to make Mailcrypt remember your passphrase for 10 minutes after each use, you would use the following line in your `.emacs' file:
(setq mc-passwd-timeout 600)
A value of nil
or 0 will disable passphrase caching completely.
This provides some increase in security, but be aware that you are
already playing a dangerous game by typing your passphrase at a Lisp
interpreter.
Mailcrypt understands multiple secret keys with distinct passphrases.
To manually force Mailcrypt to forget your passphrase(s), use the
function mc-deactivate-passwd
. Both mc-read-mode
and
mc-write-mode
bind this function to C-c / f by default.
Warning: Although Mailcrypt takes pains to overwrite your passphrase when "forgetting", it cannot prevent the Emacs garbage collector from possibly leaving copies elsewhere in memory. Also, your last 100 keystrokes can always be viewed with the function
view-lossage
, normally bound to C-h l. So be sure to type at least 100 characters after typing your passphrase if you plan to leave your terminal unattended.
Mailcrypt knows how to fetch PGP public keys from the key servers
(see section Key Servers). The function mc-pgp-fetch-key
is bound by
default to C-c / k in both mc-read-mode
and
mc-write-mode
. Additionally, mc-encrypt
,
mc-decrypt
, and mc-verify
will offer to call this function
to automatically fetch a desired key. If you call it manually, it will
prompt you for the User ID of the key to fetch.
The variable mc-pgp-fetch-methods
is a list of ways to attempt to
fetch a key. (More precisely, it is a list of functions to be called,
each of which will attempt to fetch the key.) The methods will be tried
in the order listed. The default list is:
'(mc-pgp-fetch-from-keyrings mc-pgp-fetch-from-finger mc-pgp-fetch-from-http)
For a description of these functions, see the following sections.
If you are not directly on the Internet, you probably want to obtain a copy of the global public key ring from the keyservers, install it somewhere under the name `public-keys.pgp', and do:
(setq mc-pgp-fetch-methods '(mc-pgp-fetch-from-keyrings)) (setq mc-pgp-fetch-keyring-list '("/blah/blah/blah/public-keys.pgp"))
This will allow you to fetch keys from your local copy of the global key ring instead of sending requests to the key servers directly (see section Keyring Fetch). Alternately, if your organization has a proxy HTTP server, you can configure Mailcrypt to use that. See section HTTP Fetch.
If the key is found, you will be shown the result of running PGP on it locally. This allows you to inspect the signatures on the key relative to your own keyring before you consent to having it added. Inspect the signatures carefully! Key distribution is often the Achilles' heel of public key protocols. If you blindly use keys obtained from the key servers, you are asking for trouble.
All of the methods use mc-pgp-fetch-timeout
as a timeout in
seconds; the default value is 30.
The function mc-pgp-fetch-from-keyrings
will attempt to fetch a
key from a set of keyrings on the locally accessible filesystem. This
is useful if your organization maintains a large common public keyring
whose entire contents you do not wish to duplicate on your own ring. It
is also useful if you download a copy of the global public ring from the
key servers (see section Key Servers).
The variable mc-pgp-fetch-keyring-list
controls this behavior.
It is a list of file names of public keyrings which this function will
search, in order, when seeking a key. The default value is nil
,
meaning this search will always fail.
The function mc-pgp-fetch-from-finger
will attempt to fetch a key
by fingering an address and parsing the output for a PGP public key
block.
The function mc-pgp-fetch-from-http
will attempt to fetch a key
by connecting to a key server (see section Key Servers) which has a World
Wide Web interface.
The variables mc-pgp-keyserver-address
,
mc-pgp-keyserver-port
, and mc-pgp-keyserver-url-template
control the fetching process. The default is to use Brian LaMacchia's
key server at MIT. If this default should stop working, or if you want
to help with network congestion and machine load, you can choose a
different server. As of this writing, any of the following sequences of
Emacs Lisp in your `.emacs' file will work; choose one:
;; Key server at MIT (Massachusetts, USA) ;; This is the default; these lines are only for reference ;(setq mc-pgp-keyserver-address "pgp.ai.mit.edu") ;(setq mc-pgp-keyserver-port 80) ;(setq mc-pgp-keyserver-url-template ; "/htbin/pks-extract-key.pl?op=get&search=%s")
;; Key server at UPC (Barcelona, Spain) (setq mc-pgp-keyserver-address "goliat.upc.es") (setq mc-pgp-keyserver-port 80) (setq mc-pgp-keyserver-url-template "/cgi-bin/pks-extract-key.pl?op=get&search=%s")
;; Key server at Cambridge University (Cambridge, England) (setq mc-pgp-keyserver-address "www.cl.cam.ac.uk") (setq mc-pgp-keyserver-port 80) (setq mc-pgp-keyserver-url-template "/cgi-bin/pks-extract-key.pl?op=get&search=%s")
;; Key server at UIT (Tromso, Norway) (setq mc-pgp-keyserver-address "www.service.uit.no") (setq mc-pgp-keyserver-port 80) (setq mc-pgp-keyserver-url-template "/cgi-bin/pks-extract-key.pl?op=get&search=%s")
;; Key server at CMU (Pennsylvania, USA) (setq mc-pgp-keyserver-address "gs211.sp.cs.cmu.edu") (setq mc-pgp-keyserver-port 80) (setq mc-pgp-keyserver-url-template "/cgi-bin/pgp-key?pgpid=%s")
If your organization has a firewall, you might not be able to access the World Wide Web directly. Your organization may have a proxy HTTP server set up, however. In that case, you should place code like the following in your `.emacs' file. You can use any of the above key servers instead of the one at MIT, of course.
;; Mailcrypt configuration for accessing key server through HTTP proxy (setq mc-pgp-keyserver-address "your.proxy.com") (setq mc-pgp-keyserver-port 13013) ; Your proxy's port (setq mc-pgp-keyserver-url-template "http://pgp.ai.mit.edu/htbin/pks-extract-key.pl?op=get&search=%s")
Note that fetching from a key server can be somewhat slow, so be patient. (At least it beats the tar out of the Email interface.)
This chapter documents some additional Mailcrypt configuration options which could not be naturally described elsewhere.
By default, Mailcrypt will use the same public keyring that PGP would use if executed from the shell.
You can cause Mailcrypt to use a specific public keyring by setting the
variable mc-pgp-alternate-keyring
. If this variable is set,
Mailcrypt will use that keyring for all functions which would otherwise
have used the default. This includes adding keys, extracting keys,
verifying signatures, and encrypting messages.
This feature might be useful if you maintain multiple keyrings; you can switch between them by setting this variable. Depending on your tastes, you might want to configure fetching from a keyring as well (see section Keyring Fetch).
By default, Mailcrypt will supply a "comment" option to PGP, resulting in output which looks something like this:
----- BEGIN PGP FOOBAR ----- Version: 2.6.3 Comment: Processed by Mailcrypt 3.4, an Emacs/PGP interface ... ----- END PGP FOOBAR -----
To change the comment to one of your own, set the variable
mc-pgp-comment
. Set it to nil
to use PGP's default, which
is probably either no comment or something defined in `config.txt'.
mc-read-mode
and mc-write-mode
will each indicate they are
active by placing the string `MC-r' or `MC-w' in the mode
line, respectively.
You can change these strings by setting the variables
mc-read-mode-string
and mc-write-mode-string
. So, for
example, to get rid of the mode indicators entirely, you might put the
following lines into your `.emacs' file:
(setq mc-read-mode-string "") (setq mc-write-mode-string "")
The Mailcrypt key bindings are defined by the keymaps
mc-read-mode-map
and mc-write-mode-map
. To change the key
bindings, you just need to set these variables in your `.emacs'
file.
For example, if you wanted C-c C-m to be the Mailcrypt prefix
(instead of C-c /) in mc-read-mode
, you would put the
following code in your `.emacs' file:
(setq mc-read-mode-map (make-sparse-keymap)) (define-key mc-read-mode-map "\C-c\C-mf" 'mc-deactivate-passwd) (define-key mc-read-mode-map "\C-c\C-md" 'mc-decrypt) (define-key mc-read-mode-map "\C-c\C-mv" 'mc-verify) (define-key mc-read-mode-map "\C-c\C-ma" 'mc-snarf) (define-key mc-read-mode-map "\C-c\C-mk" 'mc-pgp-fetch-key)
For more information on Emacs key bindings, see section `Customizing Key Bindings' in The GNU Emacs Manual.
The information in this section should be unnecessary, but is provided "just in case".
Mailcrypt will look for the PGP executable in your standard search path
under the name `pgp'. To use a different name (or to provide a
complete path), set the variable mc-pgp-path
.
In order to keep your identities straight, Mailcrypt needs to know where your secret keyring resides.
Mailcrypt figures this out heuristically by assuming that the file `secring.pgp' is in the same directory as your public key ring. It determines the location of the latter by doing a dry run of PGP with `+verbose=1' and parsing the output.
If this heuristic is failing for you, you can manually tell Mailcrypt
where your secret key ring is by setting the variable
mc-pgp-keydir
, like this:
(setq mc-pgp-keydir "/users/patl/.pgp/")
Note that the trailing slash is required.
If the heuristic fails, please report it as a bug (see section Credits).
Note that if you have changed the default location of your secret
keyring, Mailcrypt will be unable to locate it. You can work around
this by either setting mc-pgp-keydir
, or by making a symbolic
link to your secret keyring from `secring.pgp' in your default
public keyring directory.
Here are some random tips.
mc-passwd-timeout
. If your paranoia permits, consider increasing
it to five or ten minutes (see section Passphrase Cache).
*MailCrypt*
buffer. Check it occasionally for status and warning messages.
Mailcrypt is a powerful program, but it is not a complete PGP interface. Perhaps some future version will be; in the meantime, you will need to use the command-line interface for some operations. Things which the current version does not support include:
*MailCrypt*
buffer; see section Tips.)
This chapter contains information and pointers to information about topics related to PGP and Mailcrypt.
Key servers are machines with a publicly accessible interface to an enormous global public keyring. Anyone may add keys to or query this keyring. Each key server holds a complete copy of the global keyring, and they arrange to keep one another informed of additions they receive.
This means you can tell any key server to add your public key to the global keyring, and all of the other servers will know about it within a day or so. Then anyone will be able to query any key server to obtain your public key.
To add your key to the keyservers, send an Email message to
pgp-public-keys@pgp.ai.mit.edu
with a subject line of `ADD'
and a body containing your public key block. With Mailcrypt installed,
you can just type C-c / x to insert your public key block
(see section Inserting a Public Key Block) into the body of the message.
For help with the Email interface to the key servers, send a message with a subject line of `HELP'. For a World Wide Web interface to the key servers, see Brian LaMacchia's home page at http://www-swiss.ai.mit.edu/~bal/.
Some other key servers include:
For a complete list, consult any good online repository of PGP information (see section Online Resources).
It is strongly recommended that you submit your key to the key servers, since many humans and programs (including Mailcrypt) may look for it there. Besides, it takes mere seconds and the pain passes quickly.
If you would like to automatically receive information about new releases of Mailcrypt, send Email to `mc-announce-request@cag.lcs.mit.edu' asking to be placed on the `mc-announce' mailing list. The mailing list is maintained manually, so please be patient.
The `mc-announce' list is reserved for announcements of new Mailcrypt versions, so it has extremely low volume. We encourage you to add yourself so we can get a rough idea of how many people are using our package.
Cryptography in general, PGP in particular, and free software are politically somewhat controversial topics. Heck, in the U.S. Congress, freedom of speech is a controversial topic. Anyway, here are some organizations you should definitely watch and preferably send lots of money.
Mailcrypt's remailer support was inspired by the Communications Decency Act of 1995 (see http://www.cdt.org/cda.html) and by the International "Church" of Scientology (see http://www.mit.edu:8001/people/rnewman/scientology/).
Mailcrypt was written by Jin Choi (jin@atype.com) and Pat LoPresti (patl@lcs.mit.edu). Please send us your bug reports and comments. Also see section Mailing List.
This documentation was mostly written by Pat LoPresti, but borrows heavily from an earlier version by Hal Abelson (hal@mit.edu).
Mailcrypt would not be as robust nor as featureful if it were not for our outstanding set of Beta testers:
This index has an entry for every key sequence, function, and variable documented in this manual.