doc/intro/intro.me

.nr DR 1	\" this is a draft copy
.nr si 3n
.he 'SENDMAIL''%'
.if \n(DR .fo '\*-DRAFT\*-'\*(td'\*-DRAFT\*-'
.ls 2
.+c
.(l C
.sz 14
SENDMAIL \*- An Internet Mail Router
.sz
.sp
Eric Allman\(dg
.sp 0.5
.i
Project INGRES
Electronics Research Lab
University of California
Berkeley, California  94720
.)l
.sp 2
.(f
This is
.if \n(DR draft
version 3.16,
last modified on 02/20/82.
.if \n(DR Please do not distribute this version without permission
.if \n(DR of the author.
.)f
.(f
\(dgAuthor's current address:
Britton-Lee, Inc.
1919 Addison Street, Suite 105.
Berkeley, California 94704.
.)f
.pp
.i Sendmail
implements a general internetwork mail routing facility,
featuring aliasing and forwarding,
automatic routing to network gateways,
and flexible configuration.
.pp
In the early days of computer networking,
the problems of identification and communication
were quite simple by today's standards.
Each node on the network
would have an address,
and resources could be identified
with a host-resource pair;
in particular,
the mail system could refer to users
using a host-username pair.
Host names and numbers had to be administered by a central authority,
but usernames could be assigned locally to each host.
One early example is the ARPANET.
However, access to the ARPANET is limited,
and the connection cost is high.
Many alternative networks appeared,
such as the Berkeley Network,
the UUCP network,
and the CHAOS network.
Each network defined its own standards
for resource identification.
.pp
As networks grew,
they eventually touched.
Certain special cases could be handled trivially
by ad hoc techniques,
as when one computer hung off another by a link,
or by providing network names that appeared local to hosts
on other networks,
as with the Ethernet at Xerox PARC.
The internet was born.
.pp
Internet topology became more complex with time.
Two networks might touch
in more than one place,
and the rapid expansion of networks
created a serious database update problem.
Since all the address syntaxes were created arbitrarily,
considerable confusion reigned.
Some networks required point-to-point routing,
which simplifies the database update problem
since only adjacent hosts must be entered
into the system tables,
while others used end-to-end routing.
Some networks used a left-associative syntax
and others used a right-associative syntax:
ambiguity raised its ugly head.
.pp
Internet proposals came to the rescue.
Basically, these proposed expanding the address pairs
to address triples,
comprised of network-host-resource.
Network numbers would be universally agreed upon,
and hosts could be assigned in the old way
on each network.
But these proposals all tended to be far-reaching
and fundamentally incompatible with the old networks.
Protocols and proposals met to do battle.
And there was the issue of assigning network numbers:
there had to be a central clearing house,
and any networks that might ever touch
had to be given their very own number.
Naturally, not everyone who thought they deserved a number
got one.
.pp
Which brings us to today.
Although it will be nice when everyone everywhere
has a unique network number,
it cannot be expected very soon.
The old, stupid networks take time to die,
and bureaucratic inertia is still the rule.
.pp
.i Sendmail
is intended to help bridge the gap
between the totally ad hoc world
of networks that know nothing of each other
and the clean, tightly-coupled world
of unique network numbers.
It uses old arbitrary address syntaxes
and resolves ambiguities using heuristics
specified by the system administrator
who creates the configuration file.
It helps guide the conversion of message formats
between disparate networks.
In short,
.i sendmail
is glue that holds the world together
until the new world is ready to be inhabited.
However, it is not unreasonable to expect
escrow to take several years to close on this
new world.
.sp
.pp
Section 1 discusses the design goals for
.i sendmail .
Section 2 gives an overview of the basic functions of the system.
In section 3,
details of usage are discussed.
A detailed description of the configuration file
is given in section 4,
including a walkthrough of a specific configuration file.
Section 5 compares
.i sendmail
to other internet mail routers,
and an evaluation of
.i sendmail
is given in section 6,
including future plans.
.sh 1 "DESIGN GOALS"
.pp
.i Sendmail
is an outgrowth of
.i delivermail,
a previous incarnation of a UNIX internetwork mail router.
.i Delivermail
was written relatively quickly.
The first version could only parse addresses based on single
characters embedded in the address,
required explicit
description of gateways,
and had only limited aliasing;
automatic forwarding of messages to another gateway and other features
came later.
.pp
Design goals for
.i delivermail
included:
.np
Compatibility with the existing mail system,
including Bell version 6 mail,
Bell version 7 mail
[UNIX80],
Berkeley
.i Mail
[Shoens79],
BerkNet mail
[Schmidt79],
and hopefully UUCP mail
[Nowitz78a, Nowitz78b].
ARPANET mail
[Crocker77a, Postel77]
was also required.
.np
Reliability, in the sense of guaranteeing
that every message is correctly delivered
or at least brought to the attention of a human
for correct disposal;
no message should ever be completely lost.
This was considered essential
because of the emphasis on mail in our environment.
This turned out to be one of the hardest goals to satisfy,
especially in the face of the many anomalous message formats
produced by various ARPANET sites.
For example,
certain sites generate incorrect from addresses
which caused error message loops.
Some hosts use blanks in names,
which created problems with
UNIX mail programs that assume that an address
is one word.
And at least one person lists his address as
.q "From: the TTY of ..." ,
giving a
.q Sender:
field with his real address.
In summary,
the obscure aspects of the ARPANET mail protocol
really
.i are
used and
are difficult to support,
but must be supported.
But even obeying the standard is insufficient.
For example,
WHARTON changes our host name from
.q BERKELEY
to
.q BERKEL- ,
which confused error processing.
Degenerate cases such as this
must be handled gracefully.
.pp
There were certain other non-goals in
.i delivermail .
These resulted from the expectation that
it would only be used at Berkeley,
and probably only at a few sites at Berkeley.
.np
It was fair game to compile configuration information
into the code,
even to assume that every host was running BerkNet.
.np
The problem of multiple gateways to a single network
was not foreseen.
For example,
all UUCP mail was sent to a single gateway host.
In fact,
Berkeley has at least three UUCP gateway hosts.
.np
No attempt was made to reduce the volume of mail across a network link
by sending only one copy of a message
to multiple recipients on the same host.
Besides the difficulty of doing this,
we failed to appreciate how much volume there would be.
For example,
one of our gateways processed a message approximately
every twenty seconds
during peak hours, many of which were duplicates.
.np
Existing software to do actual delivery
should be used whenever possible.
This resulted as much from political and practical considerations
as technical.
.pp
This resulted in an architecture illustrated in figure 1.
.(z
.hl
.ie t \
.	sp 18
.el \{\
.(c
+---------+   +---------+   +---------+
| sender1 |   | sender2 |   | sender3 |
+---------+   +---------+   +---------+
     |  	   |             |
     +----------+  +  +----------+
		|  |  |
		v  v  v
            +-------------+
            | delivermail |
            +-------------+
		|  |  |
     +----------+  +  +----------+
     |  	   |             |
     v             v             v
+---------+   +---------+   +---------+
| mailer1 |   | mailer2 |   | mailer3 |
+---------+   +---------+   +---------+
.)c
.\}

.ce
Figure 1 \*- Delivermail System Structure.
.hl
.)z
The user interacts with a mail generating and sending program.
When the mail is created,
the generator calls
.i delivermail ,
which routes the message to the correct mailer(s).
Since some of the senders may be network servers
and some of the mailers may be network users,
.i delivermail
may be used as an internet mail gateway.
.pp
.i Sendmail
maintained the goals of
.i delivermail.
Time was less of a constraint,
but not reimplementing basic mailers
has proven to be a wise move in many ways.
For example,
many internet mailers deliver local mail directly.
This is more efficient,
but builds in the design decisions
of the local mailer,
and makes it difficult to concentrate
on the
.q "real problems"
(such as locking).
Other design goals were:
.np
.i Sendmail
should operate in more complex environments,
including multiple
connections to a single network type
(such as with multiple UUCP or Ether nets
[Metcalfe76]),
requiring that the contents of an address
be considered
as well as the syntax,
in order to determine the gateway to use.
For example,
the ARPANET is bringing up a new protocol
called TCP to replace the old NCP protocol.
No host at Berkeley runs both TCP and NCP,
so it is necessary to look at the ARPANET host name
to determine whether to route mail to an NCP gateway
or a TCP gateway.
.np
Configuration should not be compiled into the code.
A single binary should be able to run as is at any site
(modulo such basic changes as the CPU type or the operating system).
We have found this seemingly unimportant goal
to be critical in real life.
Besides the simple problems that occur when any program gets recompiled
in a different environment,
many sites like to
.q fiddle
with anything that they will be recompiling anyway.
.np
.i Delivermail
only knows about one alias file
and per-user forwarding is unsupported.
Berkeley is a sufficiently relaxed environment
that the system alias file can be writable by everyone,
but other environments are not so lax.
Thus,
.i sendmail
must be able to let various groups maintain their own mailing lists,
and let individuals specify their own forwarding,
without writing the system alias file.
.np
Each user should be able to specify the mailer to execute
to process mail being delivered for them.
This allows users who are using specialized mailers
that want to use a different format to build their environment
without changing the system,
and allows specialized functions
(such as returning an
.q "I am on vacation"
message).
.np
Network traffic should be minimized
by batching addresses to a single host where possible,
without assistance by the user.
.sh 1 "OVERVIEW"
.sh 2 "System Organization"
.pp
.i Sendmail
neither interfaces with the user
nor does actual mail delivery.
Rather,
it collects a message
generated by a user interface program (UIP)
such as Berkeley
.i Mail ,
MS
[Crocker77b],
or MH
[Borden79],
edits the message as required by the destination network,
and calls appropriate mailers
to do mail delivery or queueing for network transmission\**.
.(f
\**except when mailing to a file,
when
.i sendmail
does the delivery directly.
.)f
This discipline allows the insertion of new mailers
at minimum cost.
In this sense
.i sendmail
resembles the Message Processing Module (MPM)
of [Postel79b].
.sh 2 "Operational Description"
.pp
When an agent wants to send a message,
it does a normal program call to
.i sendmail .
The arguments it passes include flags giving options
and a list of addresses of intended recipients.
It then writes the message to be sent to the standard input
of
.i sendmail .
.i Sendmail
delivers the message if possible,
saving a copy of it if there were errors,
and returns an exit status code
telling what (if anything) went wrong.
.pp
The message should have a header at the beginning.
The header is formatted as a series of lines
of the form
.(b
field-name: field-value
.)b
Field-value can be split across lines by starting the following
lines with a space or a tab.
The header is separated from the body of the message
by a blank line.
No formatting requirements are imposed on the message
except that they must be lines of text
(i.e., binary data is not allowed).
.sh 3 "Argument processing and address parsing"
.pp
The arguments to
.i sendmail
are first scanned,
and flag arguments processed.
The remaining arguments are
parsed in turn as addresses,
and a list of recipients is created.
Aliases are expanded at this step.
As much validation as possible of the addresses
is done at this step:
syntax is checked, and local addresses are verified,
but detailed checking of host names and addresses
is deferred until delivery.
Forwarding is also performed
as the local addresses are verified.
.pp
.i Sendmail
appends each address
to the recipient list after parsing.
When a name is aliased or forwarded,
the old name is retained in the list,
and a flag is set in the address header
that tells the delivery phase
to ignore this recipient.
This list is kept without duplicates,
preventing alias loops
and eliminating people receiving two copies of a message,
as might occur if a person were in two groups.
.sh 3 "Message collection"
.pp
.i Sendmail
then collects the message from the standard input.
The message header is parsed at this point.
The header is stored in memory,
and the body of the message is saved
in a temporary file.
.pp
The message is still collected even if no addresses were valid
to simplify program interface.
The message will be returned with an error.
.sh 3 "Message delivery"
.pp
For each unique mailer and host in the send list,
.i sendmail
calls the appropriate mailer.
Each mailer invocation sends to all users receiving the message on one host.
Mailers that only accept one user at a time
are handled properly.
.pp
The message is sent to the mailer
(which must read its standard input)
prepended by a customized header.
The mailer exit status code is caught and checked,
and a suitable error message given as appropriate.
The exit code must conform to a system standard
or a meaningless message
(\c
.q "Service unavailable" )
is given.
.sh 3 "Queueing for retransmission"
.pp
If the mailer returned an error status that
indicated that it might be able to handle the mail later,
.i sendmail
will queue the mail and try again later.
.sh 3 "Return to sender"
.pp
If errors occurred during processing,
.i sendmail
returns the message to the sender for retransmission.
The letter can be mailed back
or written in the file
.q dead.letter
in the sender's home directory\**.
.(f
\**Obviously, if the site giving the error is not the originating
site, the only reasonable option is to mail back to the sender.
Also, there are many more error disposition options,
but they only effect the error message \*- the
.q "return to sender"
function is always handled in one of these two ways.
.)f
.sh 2 "Configuration File"
.pp
Almost all configuration information is read at runtime
from an ASCII file,
encoding
macro definitions
(defining the value of macros used internally),
header declarations
(telling sendmail the format of header lines that it will process specially,
i.e., lines that it will add or reformat),
mailer definitions
(giving information such as the location and characteristics
of each mailer),
and address rewriting rules
(a limited production system to rewrite addresses
which is used to effectively parse the addresses).
.sh 3 Macros
.pp
Macros can be used in three ways.
Certain macros transmit
unstructured textual information
into the mail system,
such as the name
.i sendmail
will use to identify itself in error messages.
Other macros transmit information from
.i sendmail
to the configuration file
for use in creating other fields
(such as argument vectors to mailers);
e.g., the name of the sender,
and the host and user
of the recipient.
Other macros are unused internally,
and can be used as shorthand in the configuration file.
.sh 3 "Header declarations"
.pp
Header declarations inform
.i sendmail
of the format of known header lines.
Knowledge of a few header lines
is built into
.i sendmail ,
such as the
.q From:
and
.q Date:
lines.
.pp
Most configured headers
will be automatically inserted
in the outgoing message
if they don't exist in the incoming message.
Certain headers are suppressed by some mailers.
.sh 3 "Mailer declarations"
.pp
Mailer declarations tell
.i sendmail
of the various mailers available to it.
The definition specifies the internal name of the mailer,
the pathname of the program to call,
some flags associated with the mailer,
and an argument vector to be used on the call;
this vector is macro expanded before use.
.sh 3 "Address rewriting rules"
.pp
The heart of address parsing in
.i sendmail
is a set of rewriting rules.
These are an ordered list of pattern-replacement rules,
(somewhat like a production system,
except that order is critical),
which are applied to each address.
The address is rewritten textually until it is either rewritten
into a special canonical form
(i.e.,
a (mailer, host, user)
3-tuple,
such as (arpanet, usc-isif, postel)
representing the address
.q "postel@usc-isif" ),
or it falls off the end.
When a pattern matches,
the rule is reapplied until it fails.
.sh 2 "Message Header Editing"
.pp
Certain editing of the message header
occurs automatically.
Header lines can be inserted
under control of the configuration file.
Some lines can be merged;
for example,
a
.q From:
line and a
.q Full-name:
line can be merged under certain circumstances.
.sh 1 "USAGE AND IMPLEMENTATION"
.sh 2 "Arguments"
.pp
Arguments may be flags and addresses.
The flag arguments are described in Appendix A.
Following flag arguments,
address arguments may be given,
unless we are running in SMTP mode.
These follow the syntax in RFC733
[Crocker77a]
for ARPANET
address formats.
In brief, the format is:
.np
Anything in parentheses is thrown away
(as a comment).
.np
Anything in angle brackets (\c
.q "<>" )
is preferred
over anything else.
This implements the ARPANET standard that addresses of the form
.(b
username <machine-address>
.)b
will send to the electronic
.q machine-address
rather than the human
.q username.
.np
Double quotes
(\ "\ )
quote phrases;
backslashes quote characters.
Backslashes are more powerful
in that they will cause otherwise equivalent phrases
to compare differently \*- for example,
.i user
and
.i
"user"
.r
are equivalent,
but
.i \euser
is different from either of them.
.pp
The rewriting rules control remaining parsing.
(Disclaimer: some special processing is done
after rewriting local names; see below.)
Parentheses, angle brackets, and double quotes
must be properly balanced and nested.
.sh 2 "Simple Mail Transfer Protocol"
.pp
The Simple Mail Transfer Protocol
(SMTP)
[Postel81]
can be used on input by specifying the
.b \-as
flag.
This will cause
.i sendmail
to use a verbose protocol on its standard input and output
which is useful over certain types of networks.
If SMTP is used,
no addresses are passed on the command line;
these are sent over the standard input instead.
.sh 2 "Mail to Files and Programs"
.pp
Files and programs are legitimate message recipients.
Files provide archival storage of messages,
useful for project administration and history.
Programs are useful as recipients in a variety of situations,
for example,
as a public repository of systems messages
(such as the Berkeley
.i msgs
program,
or the MARS system
[Sattley78]).
.pp
Any address passing through the initial parsing algorithm
as a local address
(i.e, not appearing to be a valid address for another mailer)
is scanned for two special cases.
If prefixed by a vertical bar (\c
.q \^|\^ )
the rest of the address is processed as a shell command.
If the user name begins with a slash mark (\c
.q /\^ )
the name is used as a file name,
instead of a login name.
.pp
Files that have setuid or setgid bits set
but no execute bits set
have those bits honored if
.i sendmail
is running as root.
.sh 2 "Aliasing, Forwarding, Inclusion"
.pp
.i Sendmail
reroutes mail three ways.
Aliasing applies system wide.
Forwarding allows each user to reroute incoming mail
destined for that account.
Inclusion directs
.i sendmail
to read a file for a list of addresses,
and is normally used
in conjunction with aliasing.
.sh 3 "Aliasing"
.pp
Aliasing maps names to address lists using a system-wide file.
This file is inverted to speed access.
Only names that parse as local
are allowed as aliases;
this guarantees a unique key.
.sh 3 "Forwarding"
.pp
After aliasing,
users that are local and valid
are checked for the existence of a
.q .forward
file in their home directory.
If it exists,
the message is
.i not
sent to that user,
but rather to the list of users in that file.
The expectation is that this will normally
be one user only,
and the use will be for network mail forwarding.
.pp
Forwarding also permits a user to specify a private incoming mailer.
For example,
forwarding to:
.(b
"\^|\|/usr/local/newmail myname"
.)b
will use a different incoming mailer.
.sh 3 "Inclusion"
.pp
Inclusion is specified in ARPANET syntax:
.(b
:Include: pathname
.)b
An address of this form reads the file specified by
.i pathname
and sends to all users listed in that file.
.pp
The intent is
.i not
to support direct use of this feature,
but rather to use this as a subset of aliasing.
For example,
an alias of the form:
.(b
project: :include:/usr/project/userlist
.)b
is a method of letting a project maintain a mailing list
without interaction with the system administration,
even if the alias file is protected.
.pp
It is not necessary to rebuild the alias database
when a :include: list is changed.
.sh 2 "Message Delivery"
.pp
Internally,
the recipient list is stored as one list per mailer.
Each mailer list can be scanned trivially
and mail to each host picked out to implement message batching.
Each address is marked as it is sent,
so rescanning the list is safe;
this makes sending to mailers that can only accept one user easy.
An argument list is built as the scan proceeds.
Mail to files is detected during the scan of the send list.
.pp
When an argument vector is built,
.i sendmail
creates a pipe and subprocess for the mailer.
The parent calls an
.q "editing function"
which makes the per-mailer changes to the header
and sends the result to the mailer;
a different editing function is used for sending error messages
which prepends the error information.
.pp
The exit status from the mailer is collected
after the message is sent,
and a diagnostic is printed if appropriate.
If any mail is rejected by the mailer,
a flag is set to invoke the return-to-sender function
after all delivery completes.
.sh 2 "Exit Status"
.pp
.i Sendmail
defines a set of standard exit status codes
that should be returned by mailers.
These are in turn returned by
.i sendmail .
.sh 2 "Queued Messages"
.pp
If the mailer gave a
.q "temporary failure"
exit status,
the message is queued.
A control file is used to describe the recipients to be sent to
and various other parameters.
.sh 2 "Interaction With Other Mailers"
.pp
Two examples of how network-specific work is passed to other programs
are the incoming UUCP mailer
(\c
.i rmail )
and the outgoing ARPANET mailer.
.sh 3 "Incoming UUCP mail"
.pp
Mail coming in from the UUCP network
is not guaranteed to have a normal header line,
nor will an argument be passed telling who it is from\**.
.(f
\**As a result of this,
it is impossible to verify UUCP sender addresses.
.)f
Fortuitously,
UUCP mail calls the program
.i rmail
rather than
.i mail
or
.i sendmail .
The
.i rmail
program has been modified here to do the special-purpose parsing
necessary to decode UUCP headers
and turn them into a normal UUCP address;
this address is then passed to
.i sendmail .
.sh 3 "Outgoing ARPANET mail"
.pp
The ARPANET imposes many standards that
.i sendmail
does not care to enforce.
For example,
an arpanet sitename must be on
.i every
address,
not just the
.q "From:"
address.
Certain UNIX sites like to use
.q %
as an alternative to
.q @ ,
which must be translated.
The outgoing ARPANET mailer makes these transformations
before passing the message to the network.
.sh 1 CONFIGURATION
.pp
Configuration is controlled primarily by the file
.i /usr/lib/sendmail.cf .
.i Sendmail
should not need to be recomplied except
.np
To change operating systems
(V6, V7/32V, 4BSD).
.np
To remove or insert the DBM
(UNIX database)
library.
.np
To change ARPANET reply codes.
.np
To add headers requiring special processing.
.lp
Adding mailers or changing parsing
(i.e., rewriting)
or routing information
does not require recompilation.
.pp
If the mail is being sent by a local user,
and the file
.q .mailcf
exists in the sender's home directory,
that file is read as a configuration file
after the system configuration file.
The primary use of this is to add header lines.
This could also be used to adjust the full name by
defining the
.b x
macro; e.g.,
.(b
DxEric Allman in Outer Space
.)b
.sh 2 "Configuration File Description"
.pp
The configuration file is formatted
as a series of text lines,
each beginning with a character describing its semantics.
Blank lines and lines beginning with a sharp sign
(#)
are ignored.
Other lines are:
.(b
.ta 3n
D	define macro
H	define header
M	define mailer
S	use rewriting set
C	define word class
F	define word class from file
R	specify rewriting rule
.)b
.pp
See figure 2 for an example configuration file.
Please note that this is intended as an example only.
.(z
.hl
.sz -2
.re
##### sendmail configuration file
.sp \n(psu
### local hosts on various nets
DABerkeley
DBIngVAX
DUucbvax
.sp \n(psu
### special macros
# my name
D\&n\&MAILER-DAEMON
# UNIX header format
D\&l\&From $g  $d
# delimiter (operator) characters
D\&o\&.:@!^
# address writing style
D\&q\&$g$?x ($x)$.
.sp \n(psu
### format of headers:
H\&Date: $a
H\&From: $g$?x ($x)$.
H\&Full-Name: $x
H\&Message-Id: <$t.$p.$B@$A>
H\&Posted-Date: $a
.sp \n(psu
### name classifications
# arpanet hostnames
C\&A\&ucb berkeley
# list of local host names
C\&B\&j IngVax
# berknet hosts on the arpanet
C\&C\&i ingres ing70
# uucp hostnames
C\&U\&ucbvax ernie
.sp \n(psu
.ta \w'M\&local  'u +\w'/usr/net/bin/sendberkmail  'u +\w'rlsAmn  'u +\w'$f@$A  'u
###  mailers
M\&local	/bin/mail	rlsAmn	$f	...local\&mail -d $u
M\&prog	/bin/csh	lA	$f	...prog\&mail -fc $u
M\&berk	/usr/net/bin/sendberkmail	fxs	$B:$f	...berk\&mail -m $h -h $c -t $u
M\&arpa	/usr/lib/mailers/arpa	sAu	$f@$A	...arpa\&mail $f $h $u
M\&uucp	/usr/bin/uux	rsDxmU	$U!$f	...uucp\&mail - $h!rmail ($u)
M\&xlate	/	A	$f	/
.sp \n(psu
### rewriting rules
.ta \w'R\&CSVAX:$-h!$+u  'u +\w'$#berk$@ing70$:$+u@$+h  'u
R\&$-.$+	$1:$2	change "." to ":"
R\&$=C:$+@$-	$2@$3	delete ing70: on arpanet addresses
R\&$+@$=A	ing70:$1	delete local arpa hosts
R\&$+@$-	$#berk$@ing70$:$1@$2	send arpa mail to ing70
R\&$+^$+	$1!$2	change "^" to "!"
R\&$-!$=U!$+	csvax:$3	delete uucp loops through csvax
R\&$-!$+	csvax:$1!$2	send uucp mail to csvax
R\&$-:$-:$+	$2:$3	delete multiple berk hosts
R\&$=B:$+	$2	delete local berk hosts
R\&$-:$+	$#berk$@$1$:$2	resolve berk mail
R\&$+	$#local$:$1	resolve local mail
.sp \n(psu
### rewriting rules for from host
S\&1
R\&ing70:$+@$-	$1@$2	arpanet mail is automatic
R\&CSVAX:$-!$+	$1!$2	uucp mail is automatic
.sp \n(psu
### rewriting rules for translated sender
S\&2
R\&$-x:$-:$+	$2:$3	delete multiple berknet hosts
.sz
.sp
.ce
Figure 2.  Sample configuration file.
.hl
.)z
.sh 3 "D \*- define macro"
.(b
.b D \c
.i x\|val
.)b
.pp
This line defines a macro
with the single character name
.i x
and value
.i val .
Macros can be interpolated using the escape
.b $ \c
.i x ,
where
.i x
is the macro name.
By convention,
all upper-case letters are unused by
.i sendmail
and may be used freely by the user;
all other names are reserved for use by sendmail.
Certain macros
.i must
be defined,
and are used internally.
These are:
.(b
.ta 4n
$l	UNIX-style \*(lqFrom\*(rq line.
$n	My address in error messages.
$o	\*(lqOperators\*(rq in addresses.
$q	How to write addresses in headers.
.)b
The
.b $l
macro is expanded when
.i sendmail
wants to insert a UNIX-style
.q From
line on messages.
This typically expands to something like:
.(b
From sally  Wed Aug 12 09:15:13 1981
.)b
The
.b $n
macro is used as the name of this process
when error messages are being mailed back.
Typically,
it is wise to include an alias
so that mail to this address will be sent to root.
The
.b $o
macro defines the characters
that will separate words when addresses are being broken up.
Each of these becomes a word by itself when scanned.
Blanks and tabs are built-in separators
but are ignored,
i.e., are not turned into words.
For example, the input:
.(b
Ing70:  ZRM  @  MIT-MC  SRI-KL
.)b
Is broken up into the six words:
.(b
Ing70, :, ZRM, @, MIT-MC, SRI-KL
.)b
assuming that colon and at-sign are operators
(but hyphen is not).
The
.b $q
macro gives the format for addresses
as they should appear in headers.
This will normally be something like:
.(b
$g$?x ($x)$.
.)b
Which will give the translated from address
followed by the full name if known.
.pp
A number of macros are defined by
.i sendmail
for use as primitives.
These are:
.(l
.ta 5n
$a	The \*(lqDate:\*(rq line date in ARPANET format.
$b	The current date in ARPANET format.
$c	The hop count.
$d	The date in UNIX (ctime) format.
$f	The sender's (from) address.
$g	The sender's address translated by the mailer.
$h	The host of the recipient.
$p	The process id of sendmail in decimal.
$t	The time in seconds in decimal.
$u	The user part of the recipient.
$v	The version number of sendmail.
$x	The full name of the sender.
$y	The id of the sender's terminal.
$z	The home directory of the recipient.
.)l
.pp
There are three types of dates that can be used.
The
.b $a
and
.b $b
macros are in ARPANET format;
.b $a
is a copy of the time extracted from the
.q Date:
field of the incoming message
(if there was one),
and
.b $b
is the current date and time \*- used for postmarks.
If no
.q Date:
is found in the message,
they are the same.
The
.b $d
macro has the date in UNIX
.i ctime
format;
this is extracted from the message if possible
and is otherwise the current date.
.pp
The
.b $f
macro is the id of the sender
as originally determined;
when mailing to a specific person,
the
.b $g
macro is the address of the sender
with respect to the receiver.
For example,
if I send to
.q csvax:samwise
the
.b $f
and
.b $g
macros are:
.(b
.ta 4n
$f	eric
$g	IngVAX:eric
.)b
This only applies to the first step in the link.
For example,
sending to Ing70:drb@bbn-unix,
we have
.b $f
and
.b $g
as above for the transfer to Ing70, but:
.(b
$f	IngVAX:eric
$g	IngVAX:eric@Berkeley
.)b
for transfer to the ARPANET\**.
.(f
\**When this is actually sent to the ARPANET,
this will appear as
IngVAX.eric@Berkeley.
The translation of the colon to a period is performed
by the mailer that queues ARPANET mail.
.)f
.pp
The
.b $x
macro is set to the full name of the sender.
This can be determined in several ways.
It can be passed as a flag to
.i sendmail .
The
.q Full-Name:
line in the header is the second option,
and the comment portion of the
.q From:
line is the third.
If all of these fail,
and if the message is being originated locally,
the full name is looked up in the
.i passwd
file.
.pp
When sending, the
.b $u ,
.b $h ,
and
.b $z
macros get set to the user, host, and home directory
(respectively)
of the receiver.
The host is only set if the user is not local,
and the home directory is only set if the user is local.
.pp
The
.b $p
and
.b $t
macros are used to create unique strings.
The
.b $y
macro is set to the id of the terminal of the sender
(if known);
some systems like to put this in the
.q From
line.
The
.b $v
macro is set to the version number of
.i sendmail ,
and can be used in postmarks
to help debugging.
.pp
A primitive conditional is available during macro expansion.
The construct:
.(b
$?x text1 $: text2 $.
.)b
tests if macro
.b $ \c
.i x
is defined.
If it is,
text1 is interpolated;
otherwise,
text2 is interpolated.
.sh 3 "H \*- define header"
.(b
.b H \c
.i "Field-Name" \c
.b ":" " \c
.i "field value"
.)b
.pp
The
.b H
line looks like a regular header line,
except that the field value is macro expanded
before use.
All headers mentioned in this way
are automatically inserted
into every message
except for headers mentioned in the compile-time
configuration file
.i conf.c .
These headers are
Date,
From,
Full-Name,
Message-Id,
and
Received-Date.
To get these fields the appropriate flag
must be specified
for the receiving mailer.
.pp
Since the file
.q ".mailcf"
in the sender's home directory is read and processed,
it is possible to add customized header lines.
For example,
the .mailcf consisting of:
.(b
H\&Phone: (415) 888-7770
.)b
will add that line to every outgoing message.
.sh 3 "M \*- define mailer"
.(b F
.b M \c
.i mailer-name
.i pathname
.i flags
.i from-macro
.i "argument list"
.)b
.pp
This line is structured into fields
separated by white space (spaces or tabs).
The fields are:
.np
The internal name of the mailer,
referred to in the rewriting rules.
.np
The pathname of the program to execute for this mailer.
.np
The flags for this mailer,
described below.
.np
The macro string to become the
.b $g
macro (translated sender)
for this mailer.
.np
The argument vector passed to the mailer
(macro expanded).
.pp
The flags are a series of characters:
.ls 1
.ip f
The mailer wants a
.b \-f
.i from
flag,
but only if this is a network forward operation
(i.e.,
the mailer will give an error
if the executing user does not have special permissions).
.ip r
Same as
.b f ,
but sends a
.b \-r
flag.
.ip q
Don't print errors \*- the mailer will do it for us.
.ip S
Don't reset your userid before calling the mailer.
This would be used in a secure environment where
.i sendmail
ran as a special user.
This could be used to prevent
(or at least complicate)
forged addresses.
This option is suppressed in
.q unsafe
configuration files
(i.e., user-supplied, either on a
command line
option, or in the
.i \&.mailcf
file in the home directory).
.ip n
This mailer does not want a UNIX-style
.q From
line on the message.
.ip l
This mailer is local,
so no host will be specified.
Also,
the mailer wants special local processing
(such as a
.q Received-Date:
field).
.ip s
Strip quote characters off of addresses
before calling the mailer.
.ip m
This mailer can send to multiple users
(on the same host)
in one call.
.ip F
This mailer wants a
.q From:
header line.
.ip D
This mailer wants a
.q Date:
header line.
.ip M
This mailer wants a
.q Message-Id:
header line.
.ip x
This mailer wants a
.q Full-Name:
header line.
.ip u
Upper case should be preserved in user names.
.ip h
Upper case should be preserved in host names.
.ip e
This mailer is expensive,
and it may be desirable to limit usage.
.ip A
This mailer wants an ARPANET standard header
(equivalent to the
.b F
and
.b D
flags).
.ip U
This mailer is a UUCP mailer that wants leading from lines
of the form:
.(b
From sender <date> remote from sysname
.)b
instead of the more reasonable:
.(b
From sysname!sender <date>
.)b
A compilation flag must be on to include this code.
.ls
.lp
There should always be at least one flag,
since every message should include either a
.b x
or a
.b F
flag.
.sh 3 "S \*- use rewriting set"
.(b
.b S \c
.i N
.)b
.pp
There are three sets of rewriting rules.
Set zero is used to rewrite recipient addresses.
Set one is used to rewrite sender addresses.
Set two is applied after evaluating the
.q $g
macro,
i.e., after determining the from address for a particular mailer.
.pp
Set one can be used to eliminate implicit links.
For example,
if there exists a site on on the BerkNet called
.q Ing70
which is an ARPANET gateway,
and we are on a site called
.q IngVAX ,
ARPANET mail coming into
.q Ing70
for someone on
.q IngVAX
will read:
.(b
From: Ing70:auser@ahost
.)b
Rewriting set one can rewrite this as:
.(b
From: auser@ahost
.)b
since
.q Ing70
will be implied.
.pp
Set two is used to eliminate anomalies resulting from
forwarding.
For example,
a message received at Ing70 from mckusick on the CSVAX will
appear as:
.(b
From CSVAX:mckusick
.)b
If this is then forwarded to IngVAX,
sendmail on Ing70 will rewrite the from address as:
.(b
From Ing70:CSVAX:mckusick
.)b
The extra host reference can be eliminated by ruleset two on Ing70.
.pp
When you change to a new set,
the previous content of that set is cleared.
.sh 3 "R \*- rewriting rule"
.(b F
.b R \c
.i pattern
.i replacement
.i comments
.)b
.pp
The rewriting rules drive the address parser.
The rewriting process is essentially textual.
First,
the address to be rewritten is broken up into words.
Words are defined as strings of non-special characters
separated by white space or single special characters
as defined by the
.b $o
macro.
Then,
the words are rewritten using simple pattern matching.
Words in the pattern match themselves
unless they begin with dollar sign.
The dollar escapes have the following meanings\**:
.(f
\**These dollar escapes have nothing to do with macro expansion.
.)f
.(b
.ta 6n
$-	Match a single word.
$+	Match one or more words.
$=c	Match any word in class c (see below).
.)b
The case of letters is ignored in pattern matching
(including class comparisons).
.pp
When a pattern (also called a left hand side or LHS)
matches,
the input is rewritten as defined by the right hand side (RHS).
Acceptable escapes in the RHS are:
.(b
.ta \w'$#mailer  'u
$n	Replace from corresponding match in LHS.
$#mailer	Canonical mailer name.
$@host	Canonical host name.
$:user	Canonical user name.
.)b
The substitution from LHS to RHS is done by the index
of indefinite matches on the LHS.
Each pattern reexecutes until it fails.
As soon as the input resolves to a canonical name
(i.e.,
.q "$#mailer$@host$:user" ),
rewriting ends;
otherwise,
the next pattern is tried.
The
.q "$@host"
part is not needed
if the mailer does not require a host.
The special mailer
.q error
causes the user part to be printed as an error.
.sh 3 "C \*- define word class"
.(b F
.b C \c
.i c\|word\&1
.i word\&2 ...
.)b
.pp
There are twenty six word classes,
represented as
.q A
through
.q Z .
For example:
.(b
CVcsvax ingvax esvax
.)b
defines the words
.q csvax ,
.q ingvax ,
and
.q esvax
to all be in class
.q V ,
so that
.q $=V
on the LHS of a rewriting rule
will match any of these words.
.sh 3 "F \*- define word class from file"
.(b
.b F \c
.i c\&filename
.i format
.)b
.pp
This works analogously
to the
.b C
line except that it reads the contents of the class
from the given
.i filename .
If given,
the specified
.i format
is used as a
scanf(3)
string which should produce a single string.
.sh 2 "A Detailed Example"
.pp
We will now follow the configuration file
in figure 2
through in detail.
This example is from a version of the configuration file
for the IngVAX machine at Berkeley.
IngVAX had no interesting network connections.
Ing70 had an ARPANET connection,
and CSVAX had a UUCP connection.
All of these machines were tied together via BerkNet.
.sh 3 "Macro definitions"
.(b
DABerkeley
DBIngVAX
DUucbvax
DnMAILER-DAEMON
DlFrom $g  $d
Do.:@!^
Dq$g$?x ($x)$.
.)b
The first three macros are for convenience only,
and are used to define the local host names
on the ARPANET, BerkNet, and the UUCP net
respectively.
.pp
Macro
.b n
defines the name of
.i sendmail
when error messages are sent.
Macro
.b l
defines what the first line
of a message in UNIX format looks like,
in this case the version 7 standard of:
.(b
From sender-name  time-of-submission
.)b
The
.b o
macro
tells what characters will be distinct from names
when scanning addresses.
In this case,
dot and colon will be used
to distinguish BerkNet addresses,
at sign for ARPANET addresses,
and exclamation point and caret for UUCP addresses.
.sh 3 "Header definitions"
.(b
H\&Date: $a
H\&From: $g$?x ($x)$.
H\&Full-Name: $x
H\&Message-Id: <$t.$p.$B@$A>
H\&Posted-Date: $a
.)b
These define the headers
that may be added to a message.
The
.q Date:
is just the ARPANET idea of the date.
The
.q From:
line is the translated version of the sender,
followed by the sender's full name if known.
The
.q Full-Name:
field is used to transmit the sender's full name
when a
.q From:
line is not being sent;
these will normally be mutually exclusive.
The
.q Message-Id:
field has the time and process id's concatenated
with the BerkNet and ARPANET addresses
to make a unique string.
Finally, the
.q Posted-Date:
is the date in ARPANET format;
it differs from
.q Date:
in that it is always output as soon as the message enters
.i sendmail 's
domain,
and hence indicates the time that the message first enters
the mail delivery system
[Postel79b, NBS80].
.sh 3 "Name classifications"
.(b
C\&A\&ucb berkeley
C\&B\&j IngVax
C\&C\&i ingres ing70
C\&U\&ucbvax ernie
.)b
These commands put the words
.q ucb
and
.q berkeley
into class
.q A ,
the valid names of this site on the ARPANET.
Words
.q j
and
.q ingvax
are in class
.q B ,
the local names on BerkNet.
Class
.q C ,
the names of the site which has the ARPANET link,
has the words
.q i ,
.q ingres ,
and
.q ing70 .
Finally,
.q ucbvax
and
.q ernie
are the UUCP names of our UUCP gateway,
and are in class
.q U .
.pp
The classes will be used in the patterns of the rewriting rules
as described below.
.sh 3 "Mailer definitions"
.(b
.if n .in 0
.if t .sz -2
.ta \w'M\&local  'u +\w'/usr/net/bin/sendberkmail  'u +\w'rlsAmn  'u +\w'$f@$A  'u
M\&local	/bin/mail	rlsAmn	$f	...localmail -d $u
M\&prog	/bin/csh	lA	$f	...progmail -fc $u
M\&berk	/usr/net/bin/sendberkmail	fxs	$B:$f	...berkmail -m $h -h $c -t $u
M\&arpa	/usr/lib/mailers/arpa	sAu	$f@$A	...arpamail $f $h $u
M\&uucp	/usr/bin/uux	rsDxmU	$U!$f	...uucpmail - $h!rmail ($u)
M\&xlate	/	A	$f	/
.if n .in
.if t .sz
.)b
Six mailers are known in the configuration file.
There
.i must
be entries for local and program mail.
.pp
Local mail is sent using
/bin/mail.
It takes a
.b \-r
flag,
is local,
quote characters are stripped before sending,
takes ARPANET standard headers,
can deliver to multiple recipients at once,
and does not want a UNIX-style
.q From
line since it will add one itself.
The translated
.q from
address is the same as the raw
.q from
address,
since no network hops are made.
The argument vector has a program name,
a
.b \-d
flag (\c
.q "really deliver" ,
which must be added to /bin/mail),
and the list of recipients \*- one recipient per argument.
.pp
Mail piped through programs
is interpreted by /bin/csh.
Unlike local mail,
it does not take a
.b \-r
flag,
quotes should be left,
it can only deal with one user,
and it does want a UNIX-style
.q From
line,
but is still local and still wants an ARPANET style header.
.pp
BerkNet mail is processed by
/usr/net/bin/sendberkmail.
It takes a
.b \-f
flag,
wants a
.q Full-Name:
header line,
and wants quotes stripped.
The
.q Full-Name:
is used here because if it were given as a comment
in a
.q From:
line the machine address of the sender
would not be modified by later instantiations of
.i delivermail \**.
.(f
\**\c
.i Delivermail
did no header editing,
so
.q From:
lines were always passed untouched.
When the gateways are converted to
.i sendmail
this can be changed.
.)f
The from address as seen by the receiver is
.q IngVAX:sender ,
and it takes a flag-oriented
rather than a positional
command list.
.pp
The ARPANET wants quotes stripped,
ARPANET standard headers,
and wants the user name left with case intact.
It takes a positional command list.
.pp
UUCP mail calls
.i uux
with a
.b \-r
flag,
quotes stripped,
a
.q Date:
line,
a
.q Full-Name:
line,
and with multiple users listed.
Since UUCP is a relic of the (not so) distant past,
it requires ugly header lines.
.pp
The final entry is a dummy entry to declare the pseudo-mailer
.i xlate .
When this mailer is found
it causes a table lookup to be done in the alias file
using the host name as a key.
A
.q %s
in the value part is replaced by the user name.
The entire parsing mechanism is restarted when this is encountered.
.pp
If
.q $u
were to be missing from the argument vector for a mailer,
that mailer would be accessed using the SMTP [Postel81]
protocol.
.sh 3 "Rewriting rules for recipient addresses"
.(b
.sz -2
.ta \w'[88]  'u +\w'R\&CSVAX:$-h!$+u  'u +\w'$#berk$@ing70$:$+u@$+h  'u
[1]	R\&$-.$+	$1:$2	change "." to ":"
[2]	R\&$=C:$+@$-	$2@$3	delete ing70: on arpanet addresses
[3]	R\&$+@$=A	ing70:$1	delete local arpa hosts
[4]	R\&$+@$-	$#berk$@ing70$:$1@$2	send arpa mail to ing70
[5]	R\&$+^$+	$1!$2	change "^" to "!"
[6]	R\&$-!$=U!$+	csvax:$3	delete uucp loops through csvax
[7]	R\&$-!$+	csvax:$1!$2	send uucp mail to csvax
[8]	R\&$-:$-:$+	$2:$3	delete multiple berk hosts
[9]	R\&$=B:$+	$2	delete local berk hosts
[10]	R\&$-:$+	$#berk$@$1$:$2	resolve berk mail
[11]	R\&$+	$#local$:$1	resolve local mail
.sz
.)b
The first rule translates dots to colons.
Redundant explicit routing to the ARPANET is deleted
in the second rule.
Hops out over the ARPANET
back to us are deleted in the third rule \*-
note that the BerkNet host that we would have come in on
is inserted.
Real ARPANET mail is resolved immediately with no further ado \*-
it is sent out over the BerkNet to the ing70,
and further rewriting stops immediately.
.pp
Carets are changed to exclamation points
for UUCP addresses in the fifth rule.
The sixth rule deletes loops out into UUCP land
and back to us \*- noting that we will be left on CSVAX.
The seventh rule does forwarding of UUCP mail to the CSVAX.
Multiple BerkNet hosts are deleted in rule eight \*-
this can occur internally quite easily
as a side effect of a rewriting rule.
Rule nine deletes local BerkNet hosts.
The last two rules resolve BerkNet and local mail
by turning them into the canonical form:
.(b
$#\fInet\fP$@\fIhost\fP$:\fIuser\fP
.)b
.pp
Consider the following examples.
The numbers to the left are the rule that is being applied
to make the transformation.
.(b
.re
	esvax.asa
[1]	esvax:asa
[10]	$#berk$@esvax$:asa
.)b
.(b
	research^vax135^dmr
[5]	research!vax135^dmr
[5]	research!vax135!dmr
[7]	$#berk$@csvax$:research!vax135!dmr
.)b
.(b
	research!ucbvax!j:eric
[6]	csvax:j:eric
[8]	j:eric
[9]	eric
[11]	$#local$:eric
.)b
.(b
	ing70:wnj@Berkeley
[2]	wnj@Berkeley
[3]	ing70:wnj
[10]	$#berk$@ing70$:wnj
.)b
.sh 3 "Rewriting rules for sender addresses"
.(b
.sz -2
.ta \w'R\&CSVAX:$-h!$+u  'u +\w'$+u@$+h  'u
S\&1
R\&ing70:$+@$-	$1@$2	arpanet mail is automatic
R\&CSVAX:$-!$+	$1!$2	uucp mail is automatic
.sz
.)b
The
.b S
line starts putting the rules into set one.
These rules strip off the
.q ing70:
from incoming ARPANET mail
and the
.q CSVAX:
off of incoming UUCP mail.
.pp
The name classes could be used here,
but using literal strings is safe
because they will always be program-generated.
.sh 1 "COMPARISON WITH OTHER MAILERS"
.sh 2 "Delivermail"
.pp
.i Sendmail
is an outgrowth of
.i delivermail .
The primary differences are:
.np
Configuration information is not compiled in.
This simplifies many of the problems
of moving to other machines.
It also allows easy debugging of new mailers.
.np
Address parsing is more flexible.
For example,
.i delivermail
only supported one gateway to any network,
whereas
.i sendmail
can be sensitive to host names
and reroute to different gateways.
.np
Forwarding and
:include:
features eliminate the requirement that the system alias file
be writable by any user
(or that an update program be written,
or that the system administration make all changes).
.np
.i Sendmail
supports message batching across networks
when a message is being sent to multiple recipients.
.sh 2 "MMDF"
.pp
MMDF
[Crocker79]
spans a wider problem set than
.i sendmail .
For example,
the domain of
MMDF includes a
.q "phone network"
mailer, whereas
.i sendmail
calls on preexisting mailers in most cases.
.pp
MMDF and
.i sendmail
both support aliasing,
customized mailers,
message batching,
automatic forwarding to gateways,
queueing,
and retransmission.
MMDF supports two-stage timeout,
which
.i sendmail
does not currently support.
.sh 2 "Message Processing Module"
.pp
The Message Processing Module (MPM)
discussed by Postel [Postel79b]
matches
.i sendmail
closely in terms of its basic architecture.
However,
like MMDF,
the MPM includes the network interface software
as part of its domain.
.pp
MPM also postulates a duplex channel to the receiver,
as does MMDF.
This allows simpler handling of errors
by the mailer
than possible in
.i sendmail ;
when a message queued by
.i sendmail
is sent,
any errors must be returned to the sender
by the mailer itself.
Both MPM and MMDF mailers
can return an immediate error response,
and a single error processor can create an appropriate response.
.pp
MPM prefers passing the message as a structured message,
with type-length-value tuples.
This implies a much higher degree of cooperation
between mailers than required by
.i sendmail .
MPM also assumes a universally agreed upon internet name space
(with each address a net-host-user tuple),
which
.i sendmail
does not.
.sh 1 "EVALUATIONS AND FUTURE PLANS"
.pp
.i Sendmail
is designed to work in a nonhomogeneous environment.
Every attempt is made to avoid imposing any constraints
on the underlying mailers.
This goal has driven much of the design.
One of the major problems
has been the lack of a uniform address space,
as postulated in [Postel79a]
and [Postel79b].
.pp
A nonuniform address space implies that path will be specified
in all addresses,
either explicitly (as part of the address)
or implicitly
(as with implied forwarding to gateways).
This has the unpleasant effect of making replying to messages
exceedingly difficult,
since there is no one
.q address
for any person,
but only a way to get there from wherever you are.
.pp
Interfacing to mail programs
that were not initially intended to be applied
in an internet environment
has been amazingly successful,
and has reduced the job to a manageable task.
.pp
.i Sendmail
has knowledge of a few difficult environments
built in.
It generates ARPANET FTP compatible error messages
(prepended with three-digit numbers
[Neigus73, Postel74])
as necessary,
optionally generates UNIX-style
.q From
lines on the front of messages for some mailers,
and knows how to parse the same lines on input.
This can be inconvenient to sites which have abandoned UNIX mail,
although
.i sendmail
still adds and understands ARPANET-style
.q From:
lines.
Also,
error handling has an option customized for BerkNet.
.pp
One surprisingly major annoyance in most internet mailers
(such as MMDF)
is that the location and format of local mail is built in\**.
.(f
\**For example,
MMDF puts local mail in the file
.q .mail
\*- useful if you are running version 6.
.)f
.i Sendmail
eliminates all knowledge of location
and can function successfully with different formats.
.pp
The ability to automatically generate a response to incoming mail
(by forwarding mail to a program)
seems useful
(\c
.q "I am on vacation until late August...." )
but can create problems
such as forwarding loops
(two people on vacation whose programs send notes back and forth,
for instance)
if these programs are not well written.
A program should be written to do standard tasks correctly,
but this does not solve the general case.
It might be desirable to implement some form of load limiting.
I am unaware of any mail system that addresses this problem,
nor am I aware of any reasonable solution at this time.
.pp
.i Sendmail
should be modified to run as a daemon,
reading an MPX file
(or other IPC scheme)
to receive mail and process it.
This would reduce the cost of sending mail to writing the message
into a known file.
.i Sendmail
would be modified to have a very different argument structure.
It already has an option to read the recipients
from the message header.
A more palatable technique for giving error messages
would also have to be devised.
.pp
The configuration file is currently practically inscrutable;
considerable convenience could be realized
with a higher-level format.
For example, a description might read:
.(b
.re
(MACRO name value)
(HEADER name value
	(OPTION option) ...
	(NEEDS option) ... )
(MAILER name path xlatstring
	(OPTION option) ...
	(ARGV arg ... ))
(CLASS name word ...)
(REWRITE setname
	(RULE lhs rhs) ... )
.)b
.pp
It seems clear that common protocols will be changing soon
to accommodate changing requirements and environments.
These changes will include modifications to the message header
[NBS80]
or to the body of the message itself
(such as for multimedia messages
[Postel80]).
Other changes will include changes to communication protocols
which may effect
.i sendmail ;
for example, the changes implied by the new Mail Transfer Protocol
[Sluizer81].
These changes should be relatively trivial to integrate
into the existing system.
.pp
Many other nice features could be implemented.
For example,
if we were sure that the alias file were writable by the effective user
(i.e., if
.i sendmail
were to run setuid)
then the inverted form could be rebuilt automatically when the
text copy was changed.
However, this appears to be little more than frosting.
.pp
Some proposals call for a single address syntax,
such that the host name uniquely determines the network.
There are a number of evident problems with this.
In a large internet,
the database update problem becomes considerable,
especially under multiple managements;
this can be solved by a daemon that updates the tables
dynamically,
but it is not clear what the problems are here.
More to the point,
this requires a unique namespace among all networks.
In our current configuration
we have been unable to even find out the names of all the hosts
on the UUCP network;
to hope that on an internet with fifty or more networks
would have no name conflicts is beyond the scope of
.i sendmail .
Despite the difficulties, however,
this is probably a better long-term solution to the problem
of internet addressing.
The ambiguities implied by addresses combining
left-associative and right-associative addresses
are impossible to solve without parentheses;
acceptable for mathematical equations,
but absurd for network addresses.
.pp
A related problem occurs with the user namespace.
In tightly coupled environments,
it would be nice to have automatic forwarding between machines
on the basis of the user name alone,
without cumbersome aliases.
This would require an automatically updated database
and some method of resolving conflicts.
Ideally this would be effective even with multiple managements.
A student at Berkeley,
Alan Biocca,
is working on a facility which may provide the necessary functionality.
.pp
In the long run,
a system that understands canonical internet addresses
(net, host, user)
implemented in a world that understands these addresses
would be an incredible win.
.i Sendmail
seems to be a useful tool to pull together
the haphazard environment that exists today,
until the better tools permeate the internetwork world.
.sh 0 "ACKNOWLEDGEMENTS"
.pp
Thanks are due to Kurt Shoens for his continual cheerful
assistance and good advice,
Bill Joy for pointing me in the correct direction
(over and over),
and Mark Horton for more advice,
prodding,
and many of the good ideas.
Kurt and Eric Schmidt are to be credited
for using
.i delivermail
as a server for their programs
(\c
.i Mail
and BerkNet respectively)
before any sane person should have,
and making the necessary modifications
promptly and happily.
Eric gave me considerable advice about the perils
of network software which saved me an unknown
amount of work and grief.
Mark did the original implementation of the DBM version
of aliasing, installed the VFORK code,
wrote the current version of
.i rmail ,
and was the person who really convinced me
to put the work into
.i delivermail
to turn it into
.i sendmail .
Kurt deserves accolades for using
.i sendmail
when I was myself afraid to take the risk;
how a person can continue to be so enthusiastic
in the face of so much bitter reality is beyond me.
.pp
Kurt and Kirk McKusick
read early copies of this paper,
giving considerable useful advice.
.pp
Special thanks are reserved for Mike Stonebraker,
who knowingly allowed me to put so much work into this
when there were so many other things I really should
have been working on.
.+c
.ce
REFERENCES
.nr ii 1.5i
.ip [Borden79]
Borden, S.,
Gaines, R. S.,
and
Shapiro, N. Z.,
.ul
The MH Message Handling System: Users' Manual.
R-2367-PAF.
Rand Corporation.
October 1979.
.ip [Crocker77a]
Crocker, D. H.,
Vittal, J. J.,
Pogran, K. T.,
and
Henderson, D. A. Jr.,
.ul
Standard for the Format of ARPA Network Text Messages.
RFC 733,
NIC 41952.
In [Feinler78].
November 1977.
.ip [Crocker77b]
Crocker, D. H.,
.ul
Framework and Functions of the MS Personal Message System.
R-2134-ARPA,
Rand Corporation,
Santa Monica, California.
1977.
.ip [Crocker79]
Crocker, D. H.,
Szurkowski, E. S.,
and
Farber, D. J.,
.ul
An Internetwork Memo Distribution Facility \*- MMDF.
6th Data Communication Symposium,
Asilomar.
November 1979.
.ip [Metcalfe76]
Metcalfe, R.,
and
Boggs, D.,
.q "Ethernet: Distributed Packet Switching for Local Computer Networks" ,
.ul
Communications of the ACM 19,
7.
July 1976.
.ip [Feinler78]
Feinler, E.,
and
Postel, J.
(eds.),
.ul
ARPANET Protocol Handbook.
NIC 7104,
Network Information Center,
SRI International,
Menlo Park, California.
1978.
.ip [NBS80]
National Bureau of Standards,
.ul
Specification of a Draft Message Format Standard.
Report No. ICST/CBOS 80-2.
October 1980.
.ip [Neigus73]
Neigus, N.,
.ul
File Transfer Protocol for the ARPA Network.
RFC 542, NIC 17759.
In [Feinler78].
August, 1973.
.ip [Nowitz78a]
Nowitz, D. A.,
and
Lesk, M. E.,
.ul
A Dial-Up Network of UNIX Systems.
Bell Laboratories.
In
UNIX Programmer's Manual, Seventh Edition,
Volume 2.
August, 1978.
.ip [Nowitz78b]
Nowitz, D. A.,
.ul
Uucp Implementation Description.
Bell Laboratories.
In
UNIX Programmer's Manual, Seventh Edition,
Volume 2.
October, 1978.
.ip [Postel74]
Postel, J.,
and
Neigus, N.,
Revised FTP Reply Codes.
RFC 640, NIC 30843.
In [Feinler78].
June, 1974.
.ip [Postel77]
Postel, J.,
.ul
Mail Protocol.
NIC 29588.
In [Feinler78].
November 1977.
.ip [Postel79a]
Postel, J.,
.ul
Internet Message Protocol.
RFC 753,
IEN 85.
Network Information Center,
SRI International,
Menlo Park, California.
March 1979.
.ip [Postel79b]
Postel, J. B.,
.ul
An Internetwork Message Structure.
In
.ul
Proceedings of the Sixth Data Communications Symposium,
IEEE.
New York.
November 1979.
.ip [Postel80]
Postel, J. B.,
.ul
A Structured Format for Transmission of Multi-Media Documents.
RFC 767.
Network Information Center,
SRI International,
Menlo Park, California.
August 1980.
.ip [Postel81]
Postel, J. B.,
.ul
Simple Mail Transfer Protocol.
RFC788.
Network Information Center,
SRI International,
Menlo Park, California.
November 1981.
.ip [Schmidt79]
Schmidt, E.,
.ul
An Introduction to the Berkeley Network.
University of California, Berkeley California.
1979.
.ip [Shoens79]
Shoens, K.,
.ul
Mail Reference Manual.
University of California, Berkeley.
In UNIX Programmer's Manual,
Seventh Edition,
Volume 2C.
December 1979.
.ip [Sluizer81]
Sluizer, S.,
and
Postel, J. B.,
.ul
Mail Transfer Protocol.
RFC 780.
Network Information Center,
SRI International,
Menlo Park, California.
May 1981.
.ip [UNIX80]
.ul
The UNIX Programmer's Manual, Seventh Edition,
Virtual VAX-11 Version,
Volume 1.
Bell Laboratories,
modified by the University of California,
Berkeley California.
November 1980.
.++ A
.+c "SENDMAIL USAGE"
.pp
Arguments must be presented with flags before addresses.
The flags are:
.nr ii 1i
.ip "\-f addr"
The sender's machine address is
.i addr .
This flag is ignored unless the real user
is root,
network,
or uucp,
or if
.i addr
contains an exclamation point
(because of certain restrictions in UUCP).
.ip "\-r addr"
An obsolete form of
.b \-f .
.ip "\-h cnt"
Sets the
.q "hop count"
to
.i cnt .
This represents the number of times this message has been processed
by
.i sendmail
(to the extent that it is supported by the underlying networks).
.i Cnt
is incremented during processing,
and if it reaches
MAXHOP
(currently 30)
.i sendmail
throws away the message with an error.
.ip "\-F\&name"
Sets the full name of this user to
.i name .
.ip \-e\&p
Print error messages (default).
.ip \-e\&q
Throw away error messages.
The only response is the exit status.
.ip \-e\&m
Mail back errors.
.ip \-e\&w
.q Write
back errors \*- or mail them if the user is not logged in.
.ip \-e\&e
Do special error processing for BerkNet.
This involves mailing back the errors
but always returning a zero exit status.
.ip \-n
Don't do aliasing or forwarding.
.ip \-m
Include me in alias expansions.
Normally
.i sendmail
suppresses the sender
if in a group being sent to.
.ip \-o
Assume that header lines containing recepient names in the message
will have addresses separated by spaces
rather than by commas.
Headers are always output with commas between the names.
.ip \-i
Don't take a dot to end a message.
.ip \-t
Read the header for
.q To: ,
.q Cc: ,
and
.q Bcc:
lines, and send to everyone listed in those lists.
The
.q Bcc:
line will be deleted before sending.
Any addresses in the argument vector will be deleted
from the send list.
.ip \-a
Do special processing for the
ARPANET.
This includes reading the
.q "From:"
line from the header to find the sender,
printing
ARPANET
style messages
(preceded by three digit reply codes for compatibility with
the FTP protocol
[Neigus73, Postel74, Postel77]),
and ending lines of error messages with <CRLF>.
.ip \-a\&s
Take input over an SMTP connection on standard input and output.
This does everything the \-a flag does also.
.ip \-s
Save UNIX-style
.q From
lines at the beginning of headers.
Normally they are assumed redundant
and discarded.
.ip \-v
Give a blow-by-blow description of function.
This gives information of interest to the user
rather than for the
.i sendmail
maintainer;
for example,
aliases are printed as expanded
and mailer functions are printed as they run.
.ip \-c
If this mailer is marked as being expensive,
don't connect immediately.
This requires that queueing be compiled in,
since it will depend on a sender process to
actually send the mail.
.ip \-q
Try to execute the queued up mail.
.ip \-p
Verify as much about the addresses and message as possible
and then politely run in background.
.ip \-D
Run as a daemon.
This automatically runs SMTP.
This is not completely supported yet.
.ip \-T\&time
Set timeout interval for mail that cannot be sent.
.ip \-Q\&dir
Select directory in which mail will be queued.
Typically for debugging only.
.ip \-C\&file
Use a different configuration file.
.ip \-A\&file
Use a different alias file.
.ip \-I
Initialize the DBM version
of the alias file.
If
.b \-I
is given,
no delivery is attempted.
The DBM version will be rebuilt automatically if the DBM files
are mode 666,
or if they are owned by the effective userid.
.ip \-V
Verify the addresses only.
Only partial verification is done:
syntax is checked, and local names are verified,
but no checking normally done by the mailer is attempted.
.ip \-d\&level
Set debugging level.
.ip \-M\&x\&val
Define macro
.i x
to have value
.i val .
.nr ii 5n
.+c "OTHER CONFIGURATION"
.pp
There are some configuration changes that can be made by
recompiling
.i sendmail .
Some of these are changes to compilation flags:
.nr ii 1i
.ip V6
If set,
this will compile a version 6 system,
with 8-bit user id's,
single character tty id's,
etc.
If not set,
a version 7 system is assumed.
.ip DBM
If set,
the
.q DBM
package in UNIX is used
(see DBM(3X) in [UNIX80]).
If not set,
a much less efficient algorithm for processing aliases is used.
.ip VFORK
Set if your system has the experimental
.i vfork
system call.
See vfork(2) in [UNIX80].
If not set,
the regular
.i fork
system call is used.
This option improves performance.
.ip DEBUG
If set, debugging information is compiled in.
To actually get the debugging output,
the
.b \-d
flag must be used.
.ip LOG
If set,
the
.i syslog
routine in use at some sites is used.
This makes an informational log record
for each message processed,
and makes a higher priority log record
for internal system errors.
.ip QUEUE
This flag should be set to compile in the queueing code.
If this is not set,
mailers must accept the mail immediately
or it will be returned to the sender.
.ip SMTP
If set,
the code to handle user and server SMTP will be compiled in.
This is only necessary if your machine has some mailer
that speaks SMTP.
.ip UGLYUUCP
If you have a UUCP host adjacent to you which is not running
a reasonable version of
.i rmail ,
you will have to set this flag to include the
.q "remote from sysname"
info on the from line.
Otherwise, UUCP gets confused about where the mail came from.
.ip PARANOID
There are places where
.i sendmail
may opt for a more secure,
but probably less convenient environment.
For example,
if this flag is set
it is not possible to specify a program as an address directly;
this can only be done with an alias.
.ip NOTUNIX
If you are using a non-UNIX mail format,
you can set this flag to turn off special processing
of UNIX-style
.q "From "
lines.
.nr ii 5n
.pp
Not all header semantics are defined in the configuration file.
Header lines that should only be included by certain mailers
(as well as other more obscure semantics)
must be specified in the
.i HdrInfo
table in
.i conf.c .
This table contains the header name
(which should be in all lower case),
a set of header control flags (described below),
and a set of mailer flags,
used by some of the header flags.
The header flags are:
.nr ii \w'H_ACHECK  'u
.ip H_CHECK
Check the flags for the receiving mailer
against the third field in the
.i HdrInfo
entry.
If the mailer has any of those bits set,
send this field;
otherwise, do not send this field to that mailer.
If the field was in the message originally, however,
it will always be sent
(i.e., this only applies to headers being added by
.i sendmail ).
.ip H_ACHECK
Same as H_CHECK,
except that it even applies to headers that were in the
original message.
That is,
if this bit is set and the mailer does not have flag bits set
that intersect with the third field in this
.i HdrInfo
entry,
the header line is
.i always
deleted.
.ip H_EOH
If this header field is set,
treat it like a blank line,
i.e.,
it will signal the end of the header
and the beginning of the message text.
.ip H_FORCE
Add this header entry
even if one existed in the message before.
If a header entry does not have this bit set,
.i sendmail
will not add another header line if a header line
of this name already existed.
This would normally be used to stamp the message
by everyone who handled it.
.ip H_RCPT
If set,
this field contains recipient addresses.
This is used by the
.b \-t
flag to determine who to send to
when it is collecting recipients from the message.
.ip H_ADDR
This flag indicates that this field
contains addresses that should be rewritten
to include commas, etc.
.nr ii 5n
.lp
Let's look at a sample
.i HdrInfo
specification:
.(b
.sz -2
.ta 4n +\w'"received-from",  'u +\w'H_ADDR|H_ACHECK,  'u
struct hdrinfo	HdrInfo[] =
{
	"date",	H_CHECK,	M_NEEDDATE,
	"from",	H_CHECK,	M_NEEDFROM,
	"original-from",	H_ACHECK,	0,
	"sender",	0,	0,
	"full-name",	H_ACHECK,	M_FULLNAME,
	"to",	H_ADDR,	0,
	"cc",	H_ADDR,	0,
	"bcc",	H_ADDR|H_ACHECK,	0,
	"message-id",	H_CHECK,	M_MSGID,
	"message",	H_EOH,	0,
	"text",	H_EOH,	0,
	"received-date",	H_CHECK,	M_LOCAL,
	"received-from",	H_CHECK,	M_LOCAL,
	"via",	H_FORCE,	0,
	NULL,	0,	0,
};
.sz
.)b
This specification says that the
.q Date: ,
.q From: ,
.q Message-Id: ,
.q Received-Date: ,
and
.q Received-From:
must be requested by the mailer to be inserted.
However,
if they were in the message as received by
.i sendmail
they will be propagated.
The
.q Full-Name:
field, on the other hand,
will be deleted even if it was specified before,
unless the mailer wants it.
The
.q Original-From:
and
.q Bcc:
fields will be deleted unconditionally
(since it is never possible for a mailer's flags
to intersect with zero).
The
.q Original-From:
is in fact used internally,
and will be reinserted by ad hoc code,
but only if it differs from the
.q From:
line that would otherwise be inserted.
.q To: ,
.q Cc: ,
and
.q Bcc:
all specify recipient addresses.
The
.q Message:
and
.q Text:
fields will terminate the header;
these are specified in new protocols
[NBS80]
or used by random dissenters around the network world.
The
.q Via:
field will always be added,
and can be used to trace messages.
The
.q Sender:
field is used internally,
although no cliched special processing occurs.
.pp
There are a number of important points here.
First,
header fields are not added automatically just because they are in the
.i HdrInfo
structure;
they must be specified in the configuration file
in order to be added to the message.
Any header fields mentioned in the configuration file but not
mentioned in the
.i HdrInfo
structure have default processing performed;
that is,
they are added unless they were in the message already.
Second,
the
.i HdrInfo
structure only specifies cliched processing;
certain headers are processed specially by ad hoc code
regardless of the status specified in
.i HdrInfo .
For example,
the
.q Sender:
and
.q From:
fields are always scanned on ARPANET mail
to determine the sender;
this is used to perform the
.q "return to sender"
function.
The
.q "From:"
and
.q "Full-Name:"
fields are used to determine the full name of the sender
if possible;
this is stored in the macro
.b $x
and used in a number of ways.
Although the
.q "Original-From:"
field is specified to be deleted in
.i HdrInfo ,
it is added automatically if the
.q From:
field that would be generated internally
differs from the
.q From:
field that was specified in the message;
in this case,
the original
.q From:
field is renamed
.q Original-From: .
.pp
The file
.i conf.c
also contains the specification of ARPANET reply codes.
There are six classifications these fall into:
.(b
.sz -2
.ta \w'char  'u +\w'Arpa_Usrerr[] =  'u +\w'"888";  'u
char	Arpa_Info[] =	"050";	/* arbitrary info */
char	Arpa_Enter[] =	"350";	/* start mail input */
char	Arpa_Mmsg[] =	"256";	/* mail successful (MAIL cmd) */
char	Arpa_Fmsg[] =	"250";	/* mail successful (MLFL cmd) */
char	Arpa_Syserr[] =	"455";	/* some (transient) system error */
char	Arpa_Usrerr[] =	"450";	/* some (fatal) user error */
.sz
.)b
The class
.i Arpa_Info
is for any information that is not required by the protocol,
such as forwarding information.
.i Arpa_Enter
is output when
.i sendmail
wants to start receiving the mail.
.i Arpa_Mmsg
and
.i Arpa_Fmsg
are given if the mail is successfully delivered;
the selection of message number depends on the FTP command given
(which is communicated via the
.b \-a
flag).
.i Arpa_Syserr
is printed by the
.i syserr
routine;
typically, this occurs when something has gone wrong at the
receiving site,
with the assumption that it is a transient condition.
Finally,
.i Arpa_Usrerr
is the result of a user error
and is generated by the
.i usrerr
routine;
these are generated when the user has specified something wrong,
and hence the error is permanent,
i.e.,
it will not work simply by resubmitting the request.
.pp
If it is necessary to restrict mail through a gateway,
the
.i checkcompat
routine can be modified.
This routine is called for every recipient address.
It can return
.b TRUE
to indicate that the address is acceptable
and mail processing will continue,
or it can return
.b FALSE
to reject the recipient.
If it returns false,
it is up to
.i checkcompat
to print an error message
(using
.i usrerr )
saying why the message is rejected.
For example,
.i checkcompat
could read:
.(b
.re
.sz -2
bool
checkcompat(to)
	register ADDRESS *to;
{
	if (MsgSize > 50000 && to->q_mailer != MN_LOCAL)
	{
		usrerr("Message too large for non-local delivery");
		return (FALSE);
	}
	return (TRUE);
}
.sz
.)b
This would reject messages greater than 50000 bytes
unless they were local.
The actual use of this routine is highly dependent on the
implementation,
and use should be limited.