Network Working Group                                           N. Freed
Request For Comments: 1854                  Innosoft International, Inc.
Category: Standards Track                          A. Cargille, WG Chair
                                                           October 1995


                        SMTP Service Extension
                        for Command Pipelining

Status of this Memo

  This document specifies an Internet standards track protocol for the
  Internet community, and requests discussion and suggestions for
  improvements.  Please refer to the current edition of the "Internet
  Official Protocol Standards" (STD 1) for the standardization state
  and status of this protocol.  Distribution of this memo is unlimited.

Abstract

  This memo defines an extension to the SMTP service whereby a server
  can indicate the extent of its ability to accept multiple commands in
  a single TCP send operation. Using a single TCP send operation for
  multiple commands can improve SMTP performance significantly.

Introduction

  Although SMTP is widely and robustly deployed, certain extensions may
  nevertheless prove useful. In particular, many parts of the Internet
  make use of high latency network links.

  SMTP's intrinsic one command-one response structure is significantly
  penalized by high latency links, often to the point where the factors
  contributing to overall connection time are dominated by the time
  spent waiting for responses to individual commands (turnaround time).

  In the best of all worlds it would be possible to simply deploy SMTP
  client software that makes use of command pipelining: batching up
  multiple commands into single TCP send operations. Unfortunately, the
  original SMTP specification [1] did not explicitly state that SMTP
  servers must support this.  As a result a non-trivial number of
  Internet SMTP servers cannot adequately handle command pipelining.
  Flaws known to exist in deployed servers include:

(1)   Connection handoff and buffer flushes in the middle of
      the SMTP dialogue.  Creation of server processes for
      incoming SMTP connections is a useful, obvious, and
      harmless implementation technique. However, some SMTP
      servers defer process forking and connection handoff



Freed & Cargille            Standards Track                     [Page 1]

RFC 1854                    SMTP Pipelining                 October 1995


      until some intermediate point in the SMTP dialogue.
      When this is done material read from the TCP connection
      and kept in process buffers can be lost.

(2)   Flushing the TCP input buffer when an SMTP command
      fails. SMTP commands often fail but there is no reason
      to flush the TCP input buffer when this happens.
      Nevertheless, some SMTP servers do this.

(3)   Improper processing and promulgation of SMTP command
      failures. For example, some SMTP servers will refuse to
      accept a DATA command if the last RCPT TO command
      fails, paying no attention to the success or failure of
      prior RCPT TO command results. Other servers will
      accept a DATA command even when all previous RCPT TO
      commands have failed. Although it is possible to
      accommodate this sort of behavior in a client that
      employs command pipelining, it does complicate the
      construction of the client unnecessarily.

  This memo uses the mechanism described in [2] to define an extension
  to the SMTP service whereby an SMTP server can declare that it is
  capable of handling pipelined commands. The SMTP client can then
  check for this declaration and use pipelining only when the server
  declares itself capable of handling it.

1.  Framework for the Command Pipelining Extension

  The Command Pipelining extension is defined as follows:

   (1)   the name of the SMTP service extension is Pipelining;

   (2)   the EHLO keyword value associated with the extension is
         PIPELINING;

   (3)   no parameter is used with the PIPELINING EHLO keyword;

   (4)   no additional parameters are added to either the MAIL
         FROM or RCPT TO commands.

   (5)   no additional SMTP verbs are defined by this extension;
         and,

   (6)   the next section specifies how support for the
         extension affects the behavior of a server and client
         SMTP.





Freed & Cargille            Standards Track                     [Page 2]

RFC 1854                    SMTP Pipelining                 October 1995


2.  The Pipelining Service Extension

  When a client SMTP wishes to employ command pipelining, it first
  issues the EHLO command to the server SMTP. If the server SMTP
  responds with code 250 to the EHLO command, and the response includes
  the EHLO keyword value PIPELINING, then the server SMTP has indicated
  that it can accommodate SMTP command pipelining.

2.1.  Client use of pipelining

  Once the client SMTP has confirmed that support exists for the
  pipelining extension, the client SMTP may then elect to transmit
  groups of SMTP commands in batches without waiting for a response to
  each individual command. In particular, the commands RSET, MAIL FROM,
  SEND FROM, SOML FROM, SAML FROM, and RCPT TO can all appear anywhere
  in a pipelined command group.  The EHLO, DATA, VRFY, EXPN, TURN,
  QUIT, and NOOP commands can only appear as the last command in a
  group since their success or failure produces a change of state which
  the client SMTP must accommodate. (NOOP is included in this group so
  it can be used as a synchronization point.)

  Additional commands added by other SMTP extensions may only appear as
  the last command in a group unless otherwise specified by the
  extensions that define the commands.

  The actual transfer of message content is explicitly allowed to be
  the first "command" in a group. That is, the RSET/MAIL FROM sequence
  necessary to initiate a new message transaction can be placed in the
  same group as the final transfer of the headers and body of the
  previous message.

  Client SMTP implementations that employ pipelining MUST check ALL
  statuses associated with each command in a group. For example, if
  none of the RCPT TO recipient addresses were accepted the client must
  then check the response to the DATA command -- the client cannot
  assume that the DATA command will be rejected just because none of
  the RCPT TO commands worked.  If the DATA command was properly
  rejected the client SMTP can just issue RSET, but if the DATA command
  was accepted the client SMTP should send a single dot.

  Command statuses MUST be coordinated with responses by counting each
  separate response and correlating that count with the number of
  commands known to have been issued.  Multiline responses MUST be
  supported. Matching on the basis of either the error code value or
  associated text is expressly forbidden.

  Client SMTP implementations MAY elect to operate in a nonblocking
  fashion, processing server responses immediately upon receipt, even



Freed & Cargille            Standards Track                     [Page 3]

RFC 1854                    SMTP Pipelining                 October 1995


  if there is still data pending transmission from the client's
  previous TCP send operation. If nonblocking operation is not
  supported, however, client SMTP implementations MUST also check the
  TCP window size and make sure that each group of commands fits
  entirely within the window. The window size is usually, but not
  always, 4K octets.  Failure to perform this check can lead to
  deadlock conditions.

  Clients MUST NOT confuse responses to multiple commands with
  multiline responses. Each command requires one or more lines of
  response, the last line not containing a dash between the response
  code and the response string.

2.2.  Server support of pipelining

  A server SMTP implementation that offers the pipelining extension:

   (1)   MUST NOT flush or otherwise lose the contents of the
         TCP input buffer under any circumstances whatsoever.

   (2)   SHOULD issue a positive response to the DATA command if
         and only if one or more valid RCPT TO addresses have
         been previously received.

   (3)   MUST NOT, after issuing a positive response to a DATA
         command with no valid recipients and subsequently
         receiving an empty message, send any message whatsoever
         to anybody.

   (4)   SHOULD elect to store responses to grouped RSET, MAIL
         FROM, SEND FROM, SOML FROM, SAML FROM, and RCPT TO
         commands in an internal buffer so they can sent as a
         unit.

   (5)   MUST NOT buffer responses to EHLO, DATA, VRFY, EXPN,
         TURN, QUIT, and NOOP.

   (6)   MUST NOT buffer responses to unrecognized commands.

   (7)   MUST send all pending responses immediately whenever
         the local TCP input buffer is emptied.

   (8)   MUST NOT make assumptions about commands that are yet
         to be received.

   (9)   SHOULD issue response text that indicates, either
         implicitly or explicitly, what command the response
         matches.



Freed & Cargille            Standards Track                     [Page 4]

RFC 1854                    SMTP Pipelining                 October 1995


  The overriding intent of these server requirements is to make it as
  easy as possible for servers to conform to these pipelining
  extensions.

3.  Examples

  Consider the following SMTP dialogue that does not use pipelining:

  S: <wait for open connection>
  C: <open connection to server>
  S: 220 innosoft.com SMTP service ready
  C: HELO dbc.mtview.ca.us
  S: 250 innosoft.com
  C: MAIL FROM:<[email protected]>
  S: 250 sender <[email protected]> OK
  C: RCPT TO:<[email protected]>
  S: 250 recipient <[email protected]> OK
  C: RCPT TO:<[email protected]>
  S: 250 recipient <[email protected]> OK
  C: RCPT TO:<[email protected]>
  S: 250 recipient <[email protected]> OK
  C: DATA
  S: 354 enter mail, end with line containing only "."
   ...
  C: .
  S: 250 message sent
  C: QUIT
  S: 221 goodbye

  The client waits for a server response a total of 9 times in this
  simple example. But if pipelining is employed the following dialogue
  is possible:

  S: <wait for open connection>
  C: <open connection to server>
  S: 220 innosoft.com SMTP service ready
  C: EHLO dbc.mtview.ca.us
  S: 250-innosoft.com
  S: 250 PIPELINING
  C: MAIL FROM:<[email protected]>
  C: RCPT TO:<[email protected]>
  C: RCPT TO:<[email protected]>
  C: RCPT TO:<[email protected]>
  C: DATA
  S: 250 sender <[email protected]> OK
  S: 250 recipient <[email protected]> OK
  S: 250 recipient <[email protected]> OK
  S: 250 recipient <[email protected]> OK



Freed & Cargille            Standards Track                     [Page 5]

RFC 1854                    SMTP Pipelining                 October 1995


  S: 354 enter mail, end with line containing only "."
   ...
  C: .
  C: QUIT
  S: 250 message sent
  S: 221 goodbye

  The total number of turnarounds has been reduced from 9 to 4.

  The next example illustrates one possible form of behavior when
  pipelining is used and all recipients are rejected:

  S: <wait for open connection>
  C: <open connection to server>
  S: 220 innosoft.com SMTP service ready
  C: EHLO dbc.mtview.ca.us
  S: 250-innosoft.com
  S: 250 PIPELINING
  C: MAIL FROM:<[email protected]>
  C: RCPT TO:<[email protected]>
  C: RCPT TO:<[email protected]>
  C: DATA
  S: 250 sender <[email protected]> OK
  S: 550 remote mail to <[email protected]> not allowed
  S: 550 remote mail to <[email protected]> not allowed
  S: 554 no valid recipients given
  C: QUIT
  S: 221 goodbye

  The client SMTP waits for the server 4 times here as well. If the
  server SMTP does not check for at least one valid recipient prior to
  accepting the DATA command, the following dialogue would result:

  S: <wait for open connection>
  C: <open connection to server>
  S: 220 innosoft.com SMTP service ready
  C: EHLO dbc.mtview.ca.us
  S: 250-innosoft.com
  S: 250 PIPELINING
  C: MAIL FROM:<[email protected]>
  C: RCPT TO:<[email protected]>
  C: RCPT TO:<[email protected]>
  C: DATA
  S: 250 sender <[email protected]> OK
  S: 550 remote mail to <[email protected]> not allowed
  S: 550 remote mail to <[email protected]> not allowed
  S: 354 enter mail, end with line containing only "."
  C: .



Freed & Cargille            Standards Track                     [Page 6]

RFC 1854                    SMTP Pipelining                 October 1995


  C: QUIT
  S: 554 no valid recipients
  S: 221 goodbye

4.  Security Considerations

  This RFC does not discuss security issues and is not believed to
  raise any security issues not endemic in electronic mail and present
  in fully conforming implementations of [1].

5.  Acknowledgements

  This document is based on the SMTP service extension model presented
  in RFC 1425. Marshall Rose's description of SMTP command pipelining
  in his book "The Internet Message" also served as a source of
  inspiration for this extension.

6.  References

  [1]  Postel, J., "Simple Mail Transfer Protocol", STD 10
       RFC 821, USC/Information Sciences Institute, August
       1982.

  [2]  Klensin, J., Freed, N., Rose, M., Stefferud, E.,
       and D. Crocker, "SMTP Service Extensions", RFC 1651,
       MCI, Innosoft, Dover Beach Consulting, Inc.,
       Network Management Associates, Inc., Silicon Graphics,
       Inc., July 1994.

7.  Author's Address

  Ned Freed
  Innosoft International, Inc.
  1050 East Garvey Avenue South
  West Covina, CA 91790
  USA

  Phone: +1 818 919 3600
  Fax: +1 818 919 3614
  EMail: [email protected]











Freed & Cargille            Standards Track                     [Page 7]


You are viewing proxied material from gopher.rbfh.de. The copyright of proxied material belongs to its original authors. Any comments or complaints in relation to proxied material should be directed to the original authors of the content concerned. Please see the disclaimer for more details.