<?xml version="1.0" encoding="UTF-8"?>

<!DOCTYPE rfc SYSTEM 'rfc2629.dtd' [
  <!ENTITY rfc2119 PUBLIC '' 
  'http://xml.resource.org/public/rfc/bibxml/reference.RFC.2119.xml'>
  <!ENTITY rfc5228 PUBLIC '' 
  'http://xml.resource.org/public/rfc/bibxml/reference.RFC.5228.xml'>
  <!ENTITY rfc5229 PUBLIC '' 
  'http://xml.resource.org/public/rfc/bibxml/reference.RFC.5229.xml'>
  <!ENTITY rfc5229 PUBLIC '' 
  'http://xml.resource.org/public/rfc/bibxml/reference.RFC.5229.xml'>
  <!ENTITY MANAGESIEVE PUBLIC '' 
  'reference.I-D.ietf-sieve-managesieve.xml'>
]>

<?rfc toc="yes" ?>
<?rfc symrefs="yes" ?>
<?rfc sortrefs="yes"?>
<?rfc compact="yes"?>

<rfc ipr='pre5378Trust200902' docName='draft-ietf-sieve-include-02'>
  
  <front>
    <title abbrev="Sieve Extension: Include">
      Sieve Email Filtering: Include Extension
    </title>
    
    <author initials='C.' surname='Daboo' fullname='Cyrus Daboo'>
      <organization></organization>
      <address>
        <email>cyrus@daboo.name</email>
      </address>
    </author>
    
    <author initials='A.' surname='Stone' fullname='Aaron Stone'>
      <organization></organization>
      <address>
        <email>aaron@serendipity.palo-alto.ca.us</email>
      </address>
    </author>
    
    <date month='May' year='2009' />
    <area>Applications</area>
    
    <abstract>
      <t>
      The Sieve Email Filtering "include" extension permits users to include
      one Sieve script inside another. This can make managing large scripts or
      multiple sets of scripts much easier, and allows a site and its users to
      build up libraries of scripts. Users are able to include their own personal
      scripts or site-wide scripts.
      </t>
    </abstract>
    
    <note title='Change History (to be removed prior to publication as an RFC)'>
      <t>Changes from ietf-01 to ietf-02:</t>
      <list style='letters'>
        <t>Require that script names must be constant strings, not subject to variable expansion.</t>
        <t>Try the phrase immediate script instead of current script.</t>
        <t>Clarify that "global 'varname'" and "global.varname" refer to the same variable.</t>
        <t>Drop the requirement the global keywords come after require and before anything else.</t>
      </list>
      <t>Changes from ietf-00 to ietf-01:</t>
      <list style='letters'>
        <t>Replaced import/export with global.</t>
        <t>Added :once modifier to include.</t>
        <t>Added global namespace to see if it holds water.</t>
      </list>
      <t>Changes from daboo-06 to ietf-00:</t>
      <list style='letters'>
        <t>None</t>
      </list>
      <t>Changes from -05 to -06:</t>
      <list style='letters'>
        <t>Aaron Stone joins as author.</t>
        <t>Removed | characters from the script examples.</t>
        <t>Updated draft references to published RFCs.</t>
      </list>
      <t>Changes from -04 to -05:</t>
      <list style='letters'>
        <t>Fixed examples.</t>
        <t>Relaxed requirement that imported/exported variables be set before being used.</t>
      </list>
      <t>Changes from -03 to -04:</t>
      <list style='letters'>
        <t>Fixed missing 2119 definitions.</t>
        <t>Defined interaction with variables through use of import and export commands.</t>
      </list>
      <t>Changes from -02 to -03:</t>
      <list style='letters'>
        <t>Refreshing expired draft (updated for nits).</t>
        <t>Syntax -> Usage.</t>
        <t>Updated to 3028bis reference.</t>
      </list>
      <t>Changes from -01 to -02:</t>
      <list style='letters'>
        <t>Minor formatting changes only - refreshing expired draft.</t>
      </list>
      <t>Changes from -00 to -01:</t>
      <list style='letters'>
        <t>Added IPR boiler plate.</t>
        <t>Re-ordered sections at start to conform to RFC style.</t>
        <t>Moved recursion comment into General Considerations section.</t>
        <t>Switched to using optional parameter to indicate personal vs global.</t>
        <t>Explicitly state that an error occurs when a missing script is included.</t>
      </list>
    </note>
  </front>
  
  <middle>
    
    <section title='Introduction and Overview'>
      <t>
        It's convenient to be able to break <xref target='RFC5228'>SIEVE</xref>
        scripts down into smaller components which can be reused in a variety
        of different circumstances. For example, users may want to have a
        default script and a special 'vacation' script, the latter being
        activated when the user goes on vacation. In that case the default
        actions should continue to be run, but a vacation command should be
        executed first. One option is to edit the default script to add or
        remove the vacation command as needed. Another is to have a vacation
        script that simply has a vacation command and then includes the default
        script.
      </t>
    </section>
    
    <section title='Conventions Used in This Document'>
      <t>
        The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT",
        "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this
        document are to be interpreted as described in <xref target='RFC2119'/>.
      </t>
      <t>
        Conventions for notations are as in <xref target='RFC5228'>SIEVE</xref>
        Section 1.1.
      </t>
    </section>
    
    <section title='Include Extension'>
      
      <section title='General Considerations'>
        <t>
          Sieve implementations that implement the "include", "return", and
          "global" commands described below have an identifier of "include" for
          use with the capability mechanism. If any of the "include", "return",
          or "global" commands are used in a script, the "include" capability
          MUST be listed in the "require" statement in that script.
        </t>
        
        <t>
          Sieve implementations must track the use of actions in included
          scripts so that implicit "keep" behavior can be properly determined
          based on whether any actions have executed in any script.
        </t>
        
        <t>
          Sieve implementations are allowed to limit the total number of
          nested included scripts, but MUST provide for a total of at least
          three levels of nested scripts including the top-level script. An
          error MUST be generated either when the script is uploaded to the
          Sieve repository, or when the script is executed, if any nesting
          limit is exceeded. If such an error is detected whilst processing a
          Sieve script, an implicit "keep" action MUST be executed to prevent
          loss of any messages.
        </t>
        
        <t>
          Sieve implementations MUST ensure that recursive includes are not
          possible. For example, if script "A" includes script "B", and script "B"
          includes script "A" an error MUST be generated either when the script
          is uploaded to the Sieve repository, or when the script is executed.
          If such an error is detected whilst processing a Sieve script, an
          implicit "keep" action MUST be executed to prevent loss of any
          messages.
        </t>
        
        <t>
          Sieve implementations MUST handle missing scripts being referenced
          via an includes in an existing script. An error MUST be generated
          when a missing included script is discovered during execution. If
          such an error is detected an implicit "keep" action MUST be executed
          to prevent loss of any messages.
        </t>
        
        <t>
          If the Sieve "variables" extension <xref target='RFC5229'/> is
          present, an issue arises with the "scope" of variables defined in
          scripts that may include each other. For example, if a script defines
          the variable "${status}" with one particular meaning or usage, and
          another defines "${status}" with a different meaning, then if one
          script includes the other there is an issue as to which "${status}"
          is being referenced. To solve this problem, Sieve implementations
          MUST follow the scoping rules defined in <xref target='variables'/>
          and support the "global" command defined there.
        </t>
      </section>
      
      <section title='Control Structure include'>
        <figure>
          <artwork><![CDATA[
   Usage:   include [LOCATION] [ONCE] <value: string>

            LOCATION = ":personal" / ":global"

            ONCE = ":once"
]]></artwork>
        </figure>
        <t>
          The "include" command takes an optional "location" parameter, an
          optional ":once" parameter, and a single string argument representing
          the name of the script to include for processing at that point. It is
          RECOMMENDED that implementations restrict script names according to
          <xref target='I-D.ietf-sieve-managesieve' /> Section 1.7. Implementations MUST NOT
          allow variables to be expanded into the names of Sieve scripts; in
          other words, the value MUST be a constant string as defined in
          by <xref target='RFC5229'>VARIABLES</xref> Section  3.
        </t>
        
        <t>
          The "location" parameter MUST default to ":personal" if not specified.
          The "location" has the following meanings:
        </t>
        <t></t>
        
        <list style="hanging">
          <t hangText=":personal"><vspace />
            Indicates that the named script is stored in the user's own
            personal (private) Sieve repository.
          </t>
          <t hangText=":global"><vspace />
            Indicates that the named script is stored in a site-wide Sieve
            repository, accessible to all users of the Sieve system.
          </t>
        </list>

        <t>
          The ":once" parameter tells the interpreter only to include the named
          script if it has not already been included at any other point during
          script execution. If the script has already been included, processing
          continues immediately following the include command. Implementations
          MUST NOT generate an error if an "include :once" command names a script
          whose inclusion would be recursive; in this case, the script MUST
          be considered previously included and therefore "include :once" will not
          include it again.
        </t>
          
        <t>
          Note: It is RECOMMENDED that script authors / generators use this
          parameter only when including a script that performs general duties
          such as declaring global variables and making sanity checks of the
          environment.
        </t>
        
        <t>
          The included script MUST be a valid Sieve script, including having
          necessary "require" statements for all optional capabilities used by
          the script. The scope of a "require" statement in an included script
          is for the immediate script only, not the including script. For example, if
          script "A" includes script "B", and script "B" uses the "fileinto"
          extension, script "B" must have a "require" statement for "fileinto",
          irrespective of whether script "A" has one. In addition, if script
          "A" does not have a "require" statement for "fileinto", "fileinto"
          cannot be used anywhere in script "A", even after inclusion of script
          "B".
        </t>
        
        <t>
          A "stop" command in an included script MUST stop all script
          processing, including the processing of the scripts that include the
          immediate one. The "return" command (described below) stops processing
          of the immediate script only, and allows the scripts that include it to
          continue.
        </t>
        
        <t>Examples:</t>
        
        <t>The user has four scripts stored in their personal repository:</t>
        <t></t>
        <t>"default"</t>
        <t></t>
        <list>
          <t>This is the default active script that includes several others.</t>
        </list>
        
        <artwork><![CDATA[
   require ["include"];
   
   include :personal "always_allow";
   include :global "spam_tests";
   include :personal "spam_tests";
   include :personal "mailing_lists";
]]></artwork>
        
        <t>Personal script "always_allow"</t>
        <t></t>
        <list>
          <t>This script special cases some correspondent email addresses and
            makes sure any message containing those addresses are always kept.
          </t>
        </list>
        
        <artwork><![CDATA[
   if header :is "From" "boss@example.com"
   {
       keep;
   }
   elsif header :is "From" "ceo@example.com"
   {
       keep;
   }
]]></artwork>
        
        <t>Personal script "spam_tests"</t>
        <t></t>
        <list>
          <t>
            This script does some user-specific spam tests to catch spam
            messages not caught by the site-wide spam tests.
          </t>
        </list>
        
        <artwork><![CDATA[
   require ["reject"];
   
   if header :contains "Subject" "XXXX"
   {
       reject;
   }
   elsif header :is "From" "money@example.com"
   {
       reject;
   }
]]></artwork>
        
        <t>Personal script "mailing_lists"</t>
        <t></t>
        <list>
          <t>
            This script looks for messages from different mailing lists and
            files each into a mailbox specific to the mailing list.
          </t>
        </list>
        
        <artwork><![CDATA[
   require ["fileinto"];
   
   if header :is "Sender" "owner-ietf-mta-filters@imc.org"
   {
       fileinto "lists.sieve";
   }
   elsif header :is "Sender" "owner-ietf-imapext@imc.org"
   {
       fileinto "lists.imapext";
   }
]]></artwork>
        
        <t>There is one script stored in the global repository:</t>
        
        <t>Site script "spam_tests"</t>
        <t></t>
        <list>
          <t>
            This script does some site-wide spam tests which any user at the
            site can include in their own scripts at a suitable point. The
            script content is kept up to date by the site administrator.
          </t>
        </list>
        
        <artwork><![CDATA[
   require ["reject"];
   
   if anyof (header :contains "Subject" "$$",
             header :contains "Subject" "Make money")
   {
       reject;
   }
]]></artwork>
        
        <t>
          The "include" command may appear anywhere in the script where a
          control structure is legal.
        </t>
        
        <t>Example:</t>
        
        <artwork><![CDATA[
   require ["include"];
   
   if anyof (header :contains "Subject" "$$",
             header :contains "Subject" "Make money")
   {
       include "my_reject_script";
   }
]]></artwork>
        
      </section>
      
      <section title='Control Structure return'>
        <figure>
          <artwork><![CDATA[
   Usage: return
]]></artwork>
        </figure>
        
        <t>
          The "return" command stops processing of the immediately included
          script only and returns processing control to the script which
          includes it. If used in the main script (i.e. not in an included
          script), it has the same effect as the "stop" command, including the
          appropriate "keep" action if no other actions have been executed up
          to that point.
        </t>
        
      </section>
      
      <section anchor='variables' title='Interaction with Variables'>
        <t>
          In order to avoid problems of variables in an included script
          "overwriting" those from the script that includes it, this
          specification requires that all variables defined in a script MUST be
          kept "private" to the immediate script by default - i.e. they are not
          "visible" to other scripts. This ensures that two script authors
          cannot inadvertently cause problems by choosing the same name for a
          variable.
        </t>
        
        <t>
          However, sometimes there is a need to make a variable defined in one
          script available to others. This specification defines the new
          command "global" to declare that a variable is shared among scripts.
          Effectively, two namespaces are defined: one local to the immediate
          script, and another shared among all scripts. Implementations MUST
          allow a non-global variable to have the same name as a global
          variable but have no interaction between them.
        </t>
        
        <section title='Control Structure global'>
          <figure>
            <artwork><![CDATA[
   Usage:   global <value: string-list>
]]></artwork>
          </figure>
          <t>
            The "global" command contains a string list argument that defines
            one or more names of variables to be stored in the global variable
            space.
          </t>
          
          <t>
            If a "global" command is given the name of a variable that has been
            defined in the immediate script, any value stored in the global space
            is overwritten with the value of the variable in the script.
          </t>
          
          <t>
            If a "global" command lists a variable that has not been defined in
            the global namespace, the name of the variable is nonetheless
            marked as global, and any subsequent "set" command will set the
            value of the variable in global scope.
          </t>
            
          <t>
            Interpretation of a string containing a variable marked as global,
            but without any value set, SHALL behave as any other access to an
            unknown variable, as specified in Section 3 of <xref
            target='RFC5229' /> (that is, the unknown variable reference
            evaltuates to an empty string).
          </t>
          
          <t>Example:</t>
          
          <figure title="Active script">
          <artwork><![CDATA[
   require ["variables", "include"];
   global "test";
   global "test-mailbox";

   # The included script may contain repetitive code that is
   # effectively a subroutine that can be factored out.
   set "test" "$$"
   include "spam_filter_script";

   set "test" "Make money"
   include "spam_filter_script";

   # Message will be filed according to the test that matched last.
   if string :count "${test-mailbox}" "1"
   {
       fileinto "INBOX${test-mailbox}";
       stop;
   }

   # If nothing matched, the message is implicitly kept.
]]></artwork></figure>
          
          
          <figure title="spam_filter_script">
          <artwork><![CDATA[
   require ["variables", "include"];
   global ["test", "test-mailbox"];

   if header :contains "Subject" "${test}"
   {
       set "test-mailbox" "spam-${test};
   }
]]></artwork></figure>
          
        </section>

        <section title='Variables Namespace global'>
          <t>
            In addition to the "global" command, this document  defines the
            variables namespace "global", per <xref target="RFC5229"/>, Section 3.
          </t>

          <t>Example:</t>
          <artwork><![CDATA[
   require ["variables", "include"];

   set "global.i_am_on_vacation" "1";
]]></artwork>

          <t>
            Variables declared global and variables accessed via the global
            namespace MUST be one and the same. In the following example
            script, we see the variable "i_am_on_vacation" used in a "global"
            command, and again with the "global." namespace. Consider these as
            two syntaxes with identical meaning.
          </t>

          <t>Example:</t>
          <artwork><![CDATA[
   require ["variables", "include"];
   global "i_am_on_vacation";

   set "global.i_am_on_vacation" "1";

   if string :is "${i_am_on_vacation}" "1"
   {
       vacation "It's true, I am on vacation."
   }
]]></artwork>
        
        </section>

      </section>
    </section>
    
    <section title='Security Considerations'>
      <t>
        Sieve implementations MUST ensure adequate security for the global
        script repository to prevent unauthorized changes to global scripts.
      </t>

      <t>
        Sieve implementations MUST ensure that script names are checked for
        validity and proper permissions prior to inclusion, in order to prevent
        a malicious user from gaining acess to files accessible to the mail
        server software that should not be accessible to the user.
      </t>

      <t>
        Beyond these, the "include" extension does not raise any security
        considerations that are not present in the base
        <xref target='RFC5228'>SIEVE</xref> document.
      </t>
    </section>
    
    <section anchor="iana" title="IANA Considerations">
      <t>
        The following template specifies the IANA registration of the Sieve
        extension specified in this document:
      </t>
      
      <section anchor="iana-include" title="&quot;include&quot; Extension Registration">
        <figure>
          <artwork><![CDATA[
   Capability name: include
   Description:     add the "include" command to execute other Sieve
                    scripts.
   RFC number:      this RFC
   Contact address: the Sieve discussion list <ietf-mta-filters@imc.org>
]]></artwork>
        </figure>
        
      </section>
    </section>
    
  </middle>
  
  <back>
    <references title="Normative References">
      &rfc2119;
      &rfc5228;
      &rfc5229;
    </references>
    
    <references title="Informative References">
      &MANAGESIEVE;
    </references>

    <section title="Acknowledgments">
      
      <t>
	Thanks to Ken Murchison, Rob Siemborski, Alexey Melnikov, Marc Mutz,
	Kjetil Torgrim Homme, Stephan Bosch, and Arnt Gulbrandsen for comments
	and corrections.
      </t>
      
    </section>
    
  </back>
  
</rfc>
