DataMuseum.dk

top - download

    Length: 29063 (0x7187)
    Types: TextFile
    Names: »ease.paper«

└─⟦a0efdde77⟧ Bits:30001252 EUUGD11 Tape, 1987 Spring Conference Helsinki
    └─ ⟦this⟧ »EUUGD11/euug-87hel/sec8/ease/doc/ease.paper« 

TextFile

.LP
.TL
Ease: A Configuration Language
for Sendmail
.AU
James S. Schoner
.AI
Purdue University Computing Center
West Lafayette, Indiana  47907
.sp 2
.I
.ce
ABSTRACT
.R
.PP
The rapid expansion of computer networks and ensuing variation among mailing
address formats have made address interpretation an increasingly
complex task.  In the UNIX* 4.2BSD operating system, a program named 
\fIsendmail\fR was introduced which provided a
general internetwork mail routing facility.  This facility has significantly
diminished the complexity of handling address interpretation.
.PP
\fISendmail\fR's address interpretation is based on a rewriting
system composed of
a number of rewriting rules (or productions) arranged as part of a 
configuration file.  Unfortunately, the syntactical format of a
configuration file for \fIsendmail\fR is both terse and rigid, making it
rather difficult to modify.  The standard format certainly serves its 
purpose, but, as 
the need to change these configurations increases in frequency, a more 
readable format (i.e., one that is similar to the format 
of modern programming languages) is required to permit reasonably 
quick modifications to the configuration.  As a solution to this problem, 
\fBEase\fR 
provides a level of abstraction which eliminates most of the current
syntactic hindrances
faced by programmers who must reconfigure \fIsendmail\fR's 
address parsing scheme.  
.PP
As a high-level specification format, \fBEase\fR is proving to be an 
excellent alternative to \fIsendmail\fR's cryptic 
configuration file syntax.  The syntactic structures of \fBEase\fR 
are patterned after modern language constructs, making the language
easy to learn and easy to remember.  The format of the address rewriting
rule is perhaps the most significant syntactical improvement.  It was 
undoubtedly
the most needed improvement.  Nevertheless, every element of a configuration 
file is structurally enhanced through the use of \fBEase\fR. 
.FS
*  UNIX is a trademark of AT&T Bell Laboratories.
.FE
.sp 2
.NH
Introduction
.PP
The \fBEase\fR language is a high-level specification format for \fIsendmail\fR's
configuration file.  The motivation for its development
was to fulfill a goal of providing a readable and easily modifiable 
\fIsendmail\fR configuration file format.  \fBEase\fR fulfills this goal by
shielding the programmer from the cryptic configuration specification required
by \fIsendmail\fR and providing a high-level language with which the programmer
may specify all modifications to a configuration file.  The development 
of Ease coincided with
the development of an \fBEase\fR translator, \fIet\fR,
which translates a configuration file written 
in \fBEase\fR to an
equivalent file of the standard format accepted by \fIsendmail\fR.
.NH
Ease in Profile
.PP
As will be seen in the next section, the syntax of \fBEase\fR is quite
readable and easy to learn.  In order to acquire a relevant perspective
on this issue,
the reader is advised to examine a raw configuration file for \fIsendmail\fR (the 
output
of the \fBEase\fR translator, \fIet\fR, will do nicely).  The raw syntax, while
quite fitting for quick translation, can prove to be a programmer's nightmare.  
.PP
Undoubtedly, one of the more prominent features of \fBEase\fR is the ability 
to attach
names to address fields.  When address field names are well-chosen, a distinct,
self-documenting quality becomes a visible part of the address rewriting 
rules.  Ostensibly, address field names provide a new level of semantic 
abstraction.  A brief comparison of the formats can be accomplished by examining
the following equivalent representations of an address pattern:
.DS
	user_path@host_name			(\fBEase\fR format)
	$+@$-					(raw format)
.DE
In the above, \*Quser_path\*U represents a field of one or more address
tokens, and \*Qhost_name\*U represents one address token exactly.  These
token fields are represented by \*Q$+\*U and \*Q$-\*U in the raw format.  Clearly, 
the \fBEase\fR format is preferable, not only for increased readability, but 
structural comprehension as well.
.PP
Other features of \fBEase\fR include ruleset naming, long identifiers for 
macros and classes, flow-of-control structures, and free formatting.  In
addition, the C language preprocessor (cpp) can be used for file inclusion
and conditionally defined code constructs.  The next section describes
the \fBEase\fR language in complete detail.
.NH
Ease Syntax*
.FS
*  \fINo attempt is made to describe the complete semantic meaning 
associated with all of the constructs of a sendmail configuration file.  Items 
not covered in this document include the semantic distinction among rulesets, 
the special uses of
pre-defined macros, and the method of building configuration files.  To
obtain this information, the reader is advised to refer to
the Installation and Operation Guide for Sendmail (UNIX
Programmer's Manual, Volume 2c), by Eric Allman.\fR
.FE
.PP
At its highest level, \fBEase\fR can be viewed as a collection of 
block-structures, where each block begins with a keyword and is followed by
zero or more related definitions and/or declarations.  There are ten distinct 
block types.  The following is 
a list containing all ten block keywords and the block type it denotes.
.TS
center;
l l .
\fIbind\fR	-ruleset identifier bindings
\fImacro\fR	-macro definitions
\fIclass\fR	-class definitions
\fIoptions\fR	-\fIsendmail\fR option definitions
\fIprecedence\fR	-precedence definitions
\fItrusted\fR	-trusted users
\fIheader\fR	-mail header definitions
\fImailer\fR	-mailer definitions
\fIfield\fR	-address field definitions
\fIruleset\fR	-address rewriting rules
.TE
.sp 1
In general,
.TS
center ;
l .

* Letters are distinguished by case,

T{
* An \fBEase\fR identifier is defined to be a letter, followed by zero or 
more letters, digits, underscores (_), or dashes (-),
T}

T{
* A literal newline or double quotation (") character may be included in 
any quoted string by preceding the character with a backslash (\\\\\), and
T}

T{
* \fBEase\fR source is preprocessed by the C language preprocessor (cpp),
thus source comments (i.e., text enclosed by \*Q/*\*U and \*Q*/\*U) may appear 
anywhere as part of \fBEase\fR whitespace.
T}
.TE
.PP
For notational convenience, this document specifies all reserved
words of the \fBEase\fR language in italics.  In addition, quantities
enclosed in angle brackets (<..>) represent arbitrary 
identifiers, strings, or numbers.  
.NH 2
Ruleset Identifier Bindings
.PP
A ruleset (a set of rewriting rules) is identified solely by an integer 
in \fIsendmail\fR's
configuration file.  \fBEase\fR, however, allows each ruleset to be named with
a meaningful identifier.  Since a special numeric association for each 
ruleset is required by the address parsing scheme of \fIsendmail\fR, a \fIbind\fR
block must be present in any \fBEase\fR file which defines one or more 
rulesets.  A
\fIbind\fR block consists of the keyword \fIbind\fR, followed by zero or more
statements of the form:
.TS
center box;
l .
<ruleset-id> = \fIruleset\fR <ruleset-number> ;
.TE
The following example, 
.sp 1
\fIbind\fR
.PP
FINAL_RW = \fIruleset\fR 4;
.sp 1
specifies that FINAL_RW, the final rewriting ruleset, is \fIsendmail\fR's ruleset 
number 4.
.NH 2
Macro Definitions
.PP
A macro is an identifier which, when referenced in the text of a program,
is replaced by its value, a string of zero or more characters.  The value
of a macro may include references to other macros, but not itself!  \fISendmail\fR
allows a maximum of 26 user-declared macros in its configuration file.  In 
addition, there are a number of pre-declared macros which have special meaning
to \fIsendmail\fR (see Appendix A).  \fBEase\fR macros are defined in 
\fImacro\fR blocks.  \fBEase\fR allows any macro to be declared 
(which is equivalent to simply referencing it) before it is defined.  A macro
identifier is replaced by its value when it is preceded by the character
\*Q$\*U.  In addition, a macro reference inside a quoted string must always 
include braces ({}) around the macro identifier (for delimiting purposes).  
.PP
A \fImacro\fR block consists of the keyword \fImacro\fR, followed by zero
or more statements taking either of the following forms:
.TS
center box;
l .
<macro-identifier> = "<macro-value>" ;
<macro-identifier> = \fBconditional-expression\fR ;
.TE
The \fBconditional-expression\fR format will be discussed 
later.  
.sp 1
The following example,
.sp 1
\fImacro\fR
.PP
first_name = "James";
.PP
last_name = "Schoner";
.PP
whole_name = "${first_name} ${second_name}";
.sp 1
defines the macros first_name, last_name, and whole_name, where whole_name
is the string, "James Schoner".
.NH 2
Class definitions
.PP
A class is denoted by an identifier representing a logical grouping of zero 
or more names.  Classes are used to represent the range of values a token
may assume in the pattern matching of an address.  Further discussion on the
use of classes will be deferred until address fields are described.
.PP
One identifier may be used to distinctly represent both a macro
and class (i.e., the set of macro identifiers and the set of class identifiers
may form a non-empty intersection).  A name, or class element, may 
be an identifier or any quoted word.
.PP
A \fIclass\fR block consists of the keyword \fIclass\fR, followed by zero
or more statements taking any of the following forms:
.TS
center box;
l .
<class-identifier> = { <name1>, <name2>, <name3>, ... } ;
<class-identifier> = \fIreadclass\fR ( "<file-name>" ) ;
<class-identifier> = \fIreadclass\fR ( "<file-name>", "<read-format>" ) ;
.TE
The second and third forms cause \fIsendmail\fR to read the names of the class 
from the named
file.  The third form contains a read format, which should be a \fIscanf(3)\fR 
pattern yielding a single string.
.sp 1
The following example,
.sp 1
\fIclass\fR
.PP
campus_hosts = { statistics, engineering, chemistry, physics, physics-2 } ;
.PP
versions     = { "1.0", "1.1", "4.0", "4.2", latest-and-greatest } ;
.PP
phone_hosts  = \fIreadclass\fR ( "/tmp/phonenet.list" ) ;
.sp 1
defines the classes campus_hosts, versions, and phone_hosts.
.NH 2
Sendmail option definitions
.PP
A number of options to the \fIsendmail\fR program may be specified in 
an \fIoptions\fR
block.  For a description of the various \fIsendmail\fR options and their 
values, see Appendix B.  
.PP
An
\fIoptions\fR block consists of the keyword \fIoptions\fR, followed by zero
or more statements taking any of the following forms:
.TS
center box;
l l .
<option-identifier>	= "<option-value>" ;
\fIo_delivery\fR	= \fBspecial-value\fR ;
\fIo_handling\fR	= \fBspecial-value\fR ;
.TE
All but two options (\fIo_delivery\fR and \fIo_handling\fR) use the first 
form.  To specify an option without a value, simply assign to it the null 
string ("").  The \fBspecial-value\fR field of the second and third form
refers to special values (non-quoted) which are specified in Appendix B.
.sp 1
The following example,
.sp 1
\fIoptions\fR
.PP
\fIo_alias\fR = "/usr/lib/aliases" ;
.PP
\fIo_tmode\fR = "0600" ;
.PP
\fIo_delivery\fR = d_background ;
.sp 1
sets the options \fIo_alias\fR, \fIo_tmode\fR, and \fIo_delivery\fR.
.NH 2
Precedence definitions
.PP
Message headers may contain a \*QPrecedence:\*U field describing the precedence
of the message class.  Identifiers which may appear in the precedence field of
a message are given precedence values in a configuration file \fIprecedence\fR 
definition.  This association will be illustrated below in an example.
.PP
A \fIprecedence\fR block consists of the keyword \fIprecedence\fR, followed 
by zero or more statements of the form:
.KS
.TS
center box;
l .
<precedence-identifier> = <precedence-integer> ;
.TE
.KE
The following example,
.sp 1
\fIprecedence\fR
.PP
special-delivery = 100;
.PP
junk = -100;
.sp 1
defines the precedence level for the names \*Qspecial-delivery\*U and 
\*Qjunk\*U.  Thus, whenever the name \*Qjunk\*U appears in 
a \*QPrecedence:\*U field, the corresponding message class will be set to -100.
.NH 2
Trusted users
.PP
\fISendmail\fR's \fB-f\fR flag allows trusted users to override the sender's
machine address.  Trusted users are listed in \fItrusted\fR blocks.  A 
\fItrusted\fR block consists of the keyword \fItrusted\fR, followed 
by zero or more sets of users taking the form:
.TS
center box;
l .
{ <user1>, <user2>, <user3>, ... } ;
.TE
The following example,
.sp 1
\fItrusted\fR
.PP
{ root, uucp, network } ;
.PP
{ acu, kcs, jss } ;
.sp 1
specifies that the users root, uucp, network, acu, kcs, and jss can be trusted 
to use the \fIsendmail\fR flag, \fB-f\fR.
.NH 2
Mail header definitions
.PP
The format of the message headers inserted by \fIsendmail\fR is defined in one
or more \fIheader\fR blocks in the configuration file.  A \fIheader\fR block
consists of the keyword \fIheader\fR, followed by zero or more statements
taking any of the following forms:
.TS
center box;
l 
l
l
l
l
l
l
l
l
l
l .
\fIfor\fR ( <mailer-flag1>, <mailer-flag2>, ... )
       \fIdefine\fR ( "<header-title>" , \fBheader-value\fR ) ;

\fIfor\fR ( <mailer-flag1>, <mailer-flag2>, ... ) {
       \fIdefine\fR ( "<header-title>" , \fBheader-value\fR ) ;
       \fIdefine\fR ( "<header-title>" , \fBheader-value\fR ) ;
       .
       .
} ;

\fIdefine\fR ( "<header-title>" , \fBheader-value\fR ) ;
.TE
The first form is used to define one header for one or more mailer
flags.  The second form differs from the first in that more than one
header may be defined for a given set of flags.  The third form is used to 
define a header,
regardless of mailer flags.  Refer to Appendix C for a list of \fBEase\fR 
identifiers representing mailer flags.  The header title is a simple
string of characters (no macro references), whereas the \fBheader-value\fR can
be either a string of characters (possibly containing macro references) or 
a \fBconditional-expression\fR (discussed later).
.sp 1
The following example,
.DS
\fIheader\fR

	\fIdefine\fR ( "Subject:", "") ;

	\fIfor\fR ( \fIf_return\fR )
		\fIdefine\fR ( "Return-Path:", "<${\fIm_sreladdr\fR}>" ) ;

	\fIfor\fR ( \fIf_date\fR ) {
		\fIdefine\fR ( "Resent-Date:", "${\fIm_odate\fR}" ) ;
		\fIdefine\fR ( "Date:", "${\fIm_odate\fR}" );
	} ;
.DE
defines a \*QSubject\*U field for all mailers, regardless of their flags, a
\*QReturn-Path\*U field for mailers whose definition specifies
the flag, \fIf_return\fR, and the headers, \*QResent-Date\*U and \*QDate\*U,
for mailers whose definition specifies the flag, \fIf_date\fR.
.NH 2
Mailer Definitions
.PP
\fISendmail\fR's definition of a mailer (or an interface to one) occurs in a
\fImailer\fR block.  A \fImailer\fR block consists of the keyword \fImailer\fR,
followed by zero or more statements of the form:
.TS
center box;
l .
<mailer-identifier> { \fBmailer-spec\fR } ;
.TE
The field, \fBmailer-spec\fR, is a list of zero or more of the
following attribute assignments (where successive assignment statements are
separated by commas):
.TS
center ;
l l 
l l
l l
l l
l l
l l
l l .
\fIPath\fR	= \fBstring-attribute\fR
\fIArgv\fR	= \fBstring-attribute\fR
\fIEol\fR	= \fBstring-attribute\fR
\fIMaxsize\fR	= \fBstring-attribute\fR
\fIFlags\fR	= { <mailer-flag1>, <mailer-flag2>, ... } 
\fISender\fR	= <sender-ruleset-id>
\fIRecipient\fR	= <recipient-ruleset-id>
.TE
The \fBstring-attribute\fR value can take the form of a quoted string
(possibly containing macro references) or a \fBconditional-expression\fR 
(discussed later).
.sp 1
The following example,
.sp 1
\fImailer\fR
.DS
	local {
		\fIPath\fR		= "/bin/mail",
		\fIFlags\fR		= { \fIf_from\fR, \fIf_locm\fR },
		\fISender\fR	= Sender_RW,
		\fIRecipient\fR	= Recip_RW,
		\fIArgv\fR		= "mail -d ${\fIm_ruser\fR}",
		\fIMaxsize\fR	= "200000"
	} ;
.DE
defines a mailer named \*Qlocal\*U.
.NH 2
Address field definitions
.PP
\fISendmail\fR's address parsing scheme treats an address as a group of tokens
(an address token is precisely defined in the Arpanet protocol RFC822).  In
general, \fIsendmail\fR divides an address into tokens based on a list of
characters assigned as a string to the special macro \fIm_addrops\fR.  These
characters will individually be considered as tokens and will separate tokens
when parsing is performed. 
.PP
For
the \fBEase\fR language, there is a distinct set of address tokens (defined
below) which are used in combination to represent generic forms of 
addresses.  In 
addition to literal address tokens, the pattern to be matched in a rewriting 
rule (often refered to as the LHS) may
include field identifiers which match one of five possibilities:
.DS
	- zero or more tokens
	- one or more tokens
	- one token exactly
	- one token which is an element of an aribitrary class \fBX\fR
	- one token which is not an element of an aribitrary class \fBX\fR
.DE
A particular field type may be assigned to one or more identifiers.  Each
field identifier is associated with (or defined to be) a field type in
a \fIfield\fR declarations block.  A \fIfield\fR declarations block consists
of the keyword \fIfield\fR, followed by zero or more field definitions of
the form:
.TS
center box;
l .
\fBfield-id-list\fR : \fBfield-type\fR ;
.TE
A \fBfield-id-list\fR is a list of one or more identifiers, each separated by
a comma.  A \fBfield-type\fR, on the other hand, is a representation of 
one of the five fields 
described above.  The syntax for each of the five forms follows:
.DS
	\fImatch\fR ( 0* )
	\fImatch\fR ( 1* )
	\fImatch\fR ( 1 )
	\fImatch\fR ( 1 ) in <class-X>
	\fImatch\fR ( 0 ) in <class-X>
.DE
The star in the first two forms means: "or more".  Thus, the first
form would read: "match zero or more tokens".  The fourth form describes
a field where one token is matched from an arbitrary class (class-X), whereas
the fifth form describes a field where one token is matched if it is not of the
given class (class-X).
.sp 1
The following example,
.sp 1
.DS
\fIfield\fR
	anypath		: \fImatch\fR ( 0* );
	recipient_host	: \fImatch\fR ( 1 );
	local_site		: \fImatch\fR ( 1 ) in \fIm_sitename\fR;
	remote_site		: \fImatch\fR ( 0 ) in \fIm_sitename\fR;
.DE
defines the fields anypath, recipient_host, local_site, and remote_site.
.NH 2
Address rewriting rules
.PP
Address rewriting rules are grouped according to the function they perform.  For
example, it is desirable to form a distinct group for those rewriting rules 
which perform transformations on recipient addresses.
.PP
Sets of rewriting rules are defined in \fIruleset\fR blocks.  A \fIruleset\fR
block consists of the keyword \fIruleset\fR, followed by zero or more
ruleset definitions of the form:
.TS
center box;
l .
<ruleset-id> { <rewriting-rule1> <rewriting-rule2> ... }
.TE
The ruleset identifier, ruleset-id, must be defined in a \fIbind\fR block, as
described earlier.  The rewriting rules have the form:
.DS
	\fIif\fR ( <match-pattern> )
		<match-action> ( <rewriting-pattern> ) ;
.DE
where match-pattern, rewriting-pattern, and match-action are described below.
.NH 3
Match-patterns
.PP
A match-pattern is a sequence of Ease address elements representing an
address format.  If the address being rewritten matches the pattern
\*Qmatch-pattern\*U,
then the address is reformatted using the pattern \*Qrewriting-pattern\*U, and 
the corresponding
action (\*Qmatch-action\*U) is performed.  The five distinct Ease address
elements which may constitute a match-pattern are as follows:
.TS
center ;
l .
1. Field Identifiers (refer to previous section)
T{
2. Non-alphanumeric characters (the exception is the case for literal 
double quotes, which must be preceded by a backslash (\\\\\)
T}
3. Macro references
4. Quoted strings ("...")
5. \fBConditional-expressions\fR (discussed later)
.TE
Below are two sample match-patterns, each describing the same address format:
.DS
	user-id @ hostname . $arpa_suffix
	user-id @ hostname ".ARPA"
.DE
where user-id and hostname are field identifiers, and arpa_suffix is a 
user-defined macro with the value \*QARPA\*U.
.NH 3
Rewriting-patterns
.PP
A rewriting-pattern specifies the form in which to rewrite a matched 
address.  The seven distinct elements which may be used to form 
a rewriting-pattern are as follows:
.TS
center ;
l .

T{
1. Non-alphanumeric characters (the exception is the case for literal
double quotes, left parentheses, or right parentheses, each of which 
must be preceded by a backslash (\\\\\). 
T}

T{
2. A call to another ruleset.  This is used to perform rewrites
on a suffix of the rewriting-pattern.  The proper use of this
feature will be demonstrated by example below. 
T}

3. Quoted strings ("...").

4. \fBConditional-expressions\fR (discussed later).

5. A macro reference.

T{
6. A positional reference in the matched address.  A positional 
reference takes the form: $<integer-position>.  For example, 
$3 references the value of the third \fBEase\fR address 
element in the matched address.
T}

T{
7. Canonicalized host names of the form \fIcanon\fR (<id-token>),
where id-token is a regular identifier, a quoted identifier (with
double quotes), a macro reference yielding an identifier, or a 
positional reference in the matched address.  The canonicalization of 
a host name is simply a mapping to its canonical (or official) form.
T}

.TE
Below are two sample rewriting-patterns:
.DS
	$1 % $2 < @ $3 ".ARPA" >
	OLDSTYLE_RW ( $1 )
.DE
The first form specifies an address such as a%b<@c.ARPA>, where a, b, and c
represent matched identifiers or paths.  The second form specifies a call to
the ruleset \*QOLDSTYLE_RW\*U, for old-style rewriting on the parameter 
$1, which probably references the entire matched address.  This will become 
clear in later examples.
.NH 3
Match-actions
.PP
When a ruleset is called, the address to be rewritten is compared (or matched)
sequentially against the match-address of each rewriting rule.  When a
match-address describes the address \fIsendmail\fR is attempting to rewrite, the
address is rewritten (or reformatted) using the rule's 
rewriting-pattern.  Following this rewrite, the corresponding match-action
is performed.  There are four match-actions:
.TS
center ;
l l .
\fIretry\fR	T{
-a standard action which causes the rewritten address
to be again compared to the match-address of the current rule. 
T}

\fInext\fR	T{
-an action which causes the rewritten address to be
compared to the match-address of the next rewriting rule of the current 
ruleset.  If the end of the list is reached, the ruleset returns the 
rewritten address.
T}

\fIreturn\fR	T{
-an action which causes an immediate return of the 
ruleset with the current rewritten address.
T}

\fIresolve\fR	T{
-an action which specifies that the address has been
completely resolved (i.e., no further rewriting is necessary).  The 
\fIresolve\fR action is described in more detail below. 
T}
.TE
.PP
The match-action, \fIresolve\fR, is special in that it terminates
the address rewriting altogether.  The semantic structure of \fIsendmail\fR's
rewriting scheme requires that a \fIresolve\fR action appear only in the 
ruleset whose numerical binding is to the number zero.  The \fIresolve\fR action
must specify three parameters: \fImailer\fR, \fIhost\fR, and \fIuser\fR.  If
the \fImailer\fR is local, the \fIhost\fR parameter may be omitted.  The
\fImailer\fR argument must be specified as a single word, macro, or positional
reference in the matched address.  The \fIhost\fR argument may be specified as 
a single word or as an expression which expands to a single word (i.e.,
\fIhost\fR ($1 ".ARPA")).  In addition, the \fIhost\fR argument may be a
canonicalization (as described above) or a numeric internet specification.  The
keyword \fIhostnum\fR is used for numeric internet specifications, as in 
\fIhostnum\fR ("00.00.00.00") or \fIhostnum\fR ( $2 ).  The \fIuser\fR 
specification is a rewriting-pattern, as described above.  
.PP
In general, the format of a \fIresolve\fR action will be as follows:
.DS
	\fIresolve\fR (	\fImailer\fR ( <mailer-name> ),
			\fIhost\fR   ( <host-name> ),
			\fIuser\fR   ( <user-address>)   );
.DE
Examples of the match-action statement are shown below:
.DS
\fIfield\fR
	anypath	: \fImatch\fR (0*);
	usr, path	: \fImatch\fR (1*);
	hostname	: \fImatch\fR (1);
	phone_host	: \fImatch\fR (1) in phonehosts;
.DE
.DS
\fIruleset\fR

	EXAMPLE_RW {
	
		\fIif\fR ( anypath < path > anypath )   /* basic RFC821/822 parse */
			\fIretry\fR ( $2 );
		\fIif\fR ( usr " at " path )		/* \*Qat\*U -> \*Q@\*U */
			\fInext\fR ( $1 @ $2 );
		\fIif\fR ( @path: usr )
			\fIreturn\fR ( LOCAL_RW ( < @$1 > : $2 ) );
		\fIif\fR ( anypath < @phone_host".ARPA" > anypath )
			\fIresolve\fR (	\fImailer\fR ( tcp ),
					\fIhost\fR ( csnet-relay ),
					\fIuser\fR ( $1 % $2 < @"csnet-relay" > $3 ) );
	}
.DE
.PP
The example above defines the ruleset \*QEXAMPLE_RW\*U, which contains four
rewriting rules.  The first rewriting rule discards all tokens of an address
which lie on either side of a pair of angle brackets (<>), thereby 
rewriting the address as
the sequence of tokens contained within the angle brackets ($2).  Following the
address rewrite, the rule is applied again (\fIretry\fR).  When the first rule
fails to match the address being rewritten, the second rule is applied.  
.PP
The second 
rule simply replaces the word \*Qat\*U by the symbol \*Q@\*U.  The \*Q\fInext\fR\*U
action specifies that if a match is made, a rewrite is performed and 
matching continues at the next (or following) rule.  
.PP
The third rule illustrates
the use of the \*Q\fIreturn\fR\*U action, which is executed if the 
pattern \*Q@path: usr\*U
describes the current format of the address being rewritten.  In this example,
the \fIreturn\fR action returns the result of a call to ruleset \*QLOCAL_RW\*U,
which rewrites the address \*Q<@$1>:$2\*U, where $1 and $2 are substituted
with the token(s) matched respectively by \*Qpath\*U and \*Qusr\*U.
.PP
The fourth (and final) rule signals a resolution (and termination) of the
rewriting process if the given pattern is matched.  The resolution specifies
that the mailer \*Qtcp\*U will be used to deliver the message to the host
\*Qcsnet-relay\*U.  The \fIuser\fR parameter specifies the final form of the address
which \fIsendmail\fR has just resolved.
.sp 2
.PP
The \fBEase\fR construct which remains to be examined is the 
\fBconditional-expression\fR.  The \fBconditional-expression\fR provides a 
method for
constructing strings based on the condition that some test macro is (or is not)
set.  The general form begins with the concatenation of a string and a
\fBstring-conditional\fR:
.DS
	\fIconcat\fR ( <quoted-string>, \fBstring-conditional\fR )
	\fIconcat\fR ( \fBstring-conditional\fR, <quoted-string> )
.DE
A \fBstring-conditional\fR assumes either of the following forms:
.DS
	\fIifset\fR ( <macro-name>, <ifset-string> )
	\fIifset\fR ( <macro-name>, <ifset-string>, <notset-string> )
.DE
A \fBstring-conditional\fR of the first form evaluates to \*Qifset-string\*U 
if the macro \*Qmacro-name\*U has been assigned a value; otherwise it
evaluates to the null string.  The second form behaves similarly, except
that the \fBstring-conditional\fR evaluates to \*Qnotset-string\*U, instead
of the null string, if the macro \*Qmacro-name\*U has no value.
.sp 1
The following \fBconditional-expression\fR,
.DS
	\fIconcat\fR ( "New ", \fIifset\fR ( city, "York", "Jersey" ) )
.DE
evaluates to the string "New York", if the macro \*Qcity\*U is set.  Otherwise,
the \fBconditional-expression\fR evaluates to the string "New Jersey".
.NH
Ease Translation
.PP
It is important to note that \fBEase\fR is translated by a stand-alone
translator to the raw configuration file format.  No modifications were
made to the \fIsendmail\fR program itself.  As a result, syntactical verification
of a configuration file can be performed without invoking \fIsendmail\fR.
.PP
The \fBEase\fR language is translated by invoking 
the C language preprocessor (cpp) with \fBEase\fR source as input, then piping
the output as input to the \fBEase\fR translator (\fIet\fR).  The \fBEase\fR
translator may be invoked on the command line in one of four ways:
.TS
center box ;
l l .
\fIet\fR  <input-file>  <output-file>	[read from a file, write to a file]
\fIet\fR  <input-file>	[read from a file, write to standard output]
\fIet\fR  -  <output-file>	[read from standard input, write to a file]
\fIet\fR	[read from standard input, write to standard output]
.TE
.NH
Conclusion
.PP
\fBEase\fR is currently in use at the Purdue University Computing 
Center.  Source code for the \fBEase\fR translator (\fIet\fR) may be
obtained on request by writing to:
.DS
U.S. Mail:
		James S. Schoner
		c/o Kevin S. Braunsdorf
		Purdue University Computing Center
		Purdue University
		West Lafayette, Indiana  47907

Electronic Mail:
		ksb@j.cc.purdue.edu
.DE
.PP
Much of the success of this project is attributable to the constant support 
and insight offered by Mark Shoemaker.  To him, I owe a debt of gratitude.  In 
addition, I would like to thank Kevin Smallwood, Paul Albitz, and Rich Kulawiec
for their many notable suggestions and valuable insight.