1. Abstract This document describes an extension to the SMTP service [RFC1425], called Extended DATA Reply. This is an extended format of the SMTP response to the DATA verb which provides recipient-specific status information. This allows mail recipients to use individual mail filters as part of the Simple Mail Transfer Protocol. This is a very desirable feature because of escalating amounts of unwanted junk E- mail sent over the Internet. 2. Introduction The reply to the SMTP [RFC821] DATA verb is a status code indicating if the message was accepted or rejected. The status code also indicates if the rejection is temporary or permanent. Additional delivery attempts will be made for messages that are temporarily rejected. SMTP allows a message to have multiple recipients, but the reply to the DATA verb cannot indicate if the message was successfully accepted for some recipients only and rejected for the remaining recipients (permanently or temporarily). It can only indicate if the message was accepted or rejected for every one of the listed recipients. The only way to handle this condition within the current protocol is to send a status code indicating that the message was accepted, then immediately generate a non-delivery report for undeliverable recipients. This imposes additional resources and overhead to generate and transmit the non-delivery report. The ability to have recipient-specific status in the reply to the DATA verb is desirable for the following reasons: A) The receiving SMTP server does not have to generate and send a non-delivery report. B) Recipients can install individual mail filters that selectively block unwanted E-mail, and the mail filters will be able to examine the entire contents of the message before deciding to reject the message. Currently, recipient- specific filtering is only possible in response to the "RCPT TO:" verb. But the only available information at that point is the network address of the sending SMTP agent and return address of the message. Filtering unwanted junk mail based only on the network address and the return address is of very limited benefit. Expires XXX XX, XXXX [Page 1] EXDATA SMTP Extension S. Varshavchik XXX XX, XXXX There are several other mail enhancements that involve recipient- specific processing. [RFC2033] defines a related, but a completely different, mail delivery protocol which features recipient-specific processing. There was also another draft memo published, draft-ietf- fax-smtp-session, which defined an extension for recipient-specific mail processing, however its scope is different from the scope of this memo. 3. Framework for the EXDATA SMTP transport extension This SMTP transport extension [RFC1425] is laid out as follows. (1) The name of the SMTP transport extension defined here is Extended Data Reply (EXDATA). (2) The EHLO keyword associated with this extension is EXDATA. (3) The EXDATA EHLO keyword takes no parameters. (4) One optional ESMTP keyword EXDATA is associated with the MAIL FROM command. This parameter takes no values. (5) No additional ESMTP verbs are defined by this extension. (6) The next sections specify how support for this extension affects the behavior of a server and client SMTP server. 4. The EXDATA SMTP extension The receiving SMTP server indicates that it is capable of supporting the Extended Data Reply SMTP extension by including the EXDATA keyword in the EHLO list. The sending SMTP agent indicates that it supports this extension by including the EXDATA keyword in the MAIL FROM command. If the sending SMTP agent includes the EXDATA keyword in the first MAIL FROM command, it MUST include the keyword in every MAIL FROM command used in the same SMTP session. If the sending SMTP agent does not include the EXDATA keyword in the first MAIL FROM command, it MUST NOT include the keyword in any MAIL FROM command used in the same SMTP session. The receiving SMTP server SHOULD NOT send extended replies to the DATA verb if the EXDATA keyword is not included in the MAIL FROM command. An extended DATA verb reply MAY be sent instead of the second status Expires XXX XX, XXXX [Page 2] EXDATA SMTP Extension S. Varshavchik XXX XX, XXXX code reply to the DATA verb. It MUST NOT be sent as the first reply to the DATA verb (before the message is transmitted). Using the extended format in the second reply to the DATA verb is optional, and not required. The receiving SMTP server is always permitted to send regular status replies. 5. Format of an extended SMTP DATA reply The extended SMTP DATA reply uses the 558 numeric status code which is explicitly reserved for this purpose. When a 558 numeric status code is the second reply to the DATA verb, the reply MUST be formatted as follows, where , and represent the ASCII space, carriage return, and line feed characters: ex-reply-code: ex-reply-recipient-list ex-reply-recipient-list: non-final-recipient ex-reply-recipient-list | final-recipient non-final-recipient: non-term-line non-final-recipient | non-final-recipient-term-line non-term-line: "558-" result-code "-" result-text non-final-recipient-term-line: "558-" result-code result-text final-recipient: non-term-line final-recipient | final-recipient-term-line final-recipient-term-line: "558" result-code result-text result-code: digit digit digit result-text: 0 or more characters, except and , terminated by the sequence digit: Characters "0" through "9" The extended DATA reply is formatted like a standard multiline SMTP reply with a 558 numeric status code. The extended information is contained within the text portion of the 558 multiline reply. The extended information consists of one or more individual replies, where each reply itself is formatted like an SMTP reply. There will be exactly one reply for each recipient. Each individual reply may be a multiline reply itself. 6. Restrictions on the extended SMTP DATA reply Expires XXX XX, XXXX [Page 3] EXDATA SMTP Extension S. Varshavchik XXX XX, XXXX Extended replies MUST contain exactly one individual reply for each recipient whose RCPT TO command was acknowledged with a 2xx result code. Extended replies MUST NOT have any replies for recipients whose RCPT TO commands were previously rejected with a 4xx or 5xx result code. For example, if a message had three recipients, and the second RCPT TO command was rejected, the extended reply MUST contain two individual replies: the first one is for the first recipient, the second one for the last recipient. Technically, an extended SMTP DATA reply with a 2xx numeric code in every individual reply is technically equivalent to a regular 2xx reply to the DATA transaction. The "be conservative with what you send, and be liberal with what you receive" principle would dictate that the sending SMTP agent MUST be prepared to properly handle either reply, and the receiving SMTP agent SHOULD use a regular 2xx response in this situation. 7. Examples In this example the first recipient of a two-recipient message is accepted. The second recipient is rejected: 558-250 Message accepted 558-550-Access denied: 558 550 Insufficient permission In this example the first recipient is rejected. The second recipient is accepted: 558-550-Access denied 558-550 Insufficient permission 558-250-Message accepted 558 250 Queue ID is 120 7.1 Parsing extended replies Extended replies can be parsed rather easily by noting that extended replies are wrapped inside a standard SMTP multiline reply with a 558 status code. If the numeric code of the second status reply to the DATA verb is 558, and if the receiving server listed EXDATA in the EHLO keyword list, the "558-" and "558" characters are removed from the start of every line in the multiline response. The remaining text is now interpreted as a series of individual SMTP replies, one reply per recipient. Each individual reply may be a multiline reply itself. Expires XXX XX, XXXX [Page 4] EXDATA SMTP Extension S. Varshavchik XXX XX, XXXX 7.2 Parsing incomplete replies If the client receives an incomplete 558 reply, the client should process normally any of this reply’s subreplies which were received in their entirety, and should process any incomplete or unreceived subreplies as if they had been 451 subreplies. While receiving a 558 reply, the client must implement a timeout of at least 10 minutes per subreply. It MUST use per- subreply timeouts rather than trying to time the entire 558 reply. This means that the overall timeout is approximately proportional to the number of recipients. Rationale: The 10 minute limit is chosen because that’s the suggested minimum timeout for processing a message body, in the middle of section 4.5.3.2 of RFC 2821. The second and third sentences above are adapted from the first paragraph of that same section of the RFC. While sending a 558 reply, the server should take no more than about 5 minutes to send each subreply. If individual users administer their own filtering software, it would be advisable to use a timeout mechanism to force a decision after 5 minutes. Rationale: if a user’s filter always takes too long, and triggers the client’s 10 minute timeout, other users might not ever be able to receive the message. 8. Security Concerns 8.1 Transitional behavior The stated benefit of this extension is to allow recipients to use individual mail filters to reject mail during an SMTP transaction. However this extension must be widely implemented for this to happen. Furthermore, the sending SMTP agent can always omit the EXDATA keyword from the MAIL FROM: verb. It follows that recipient-specific mail filtering is still a very desireable feature if the sending SMTP agent does not state its support for the EXDATA extension. One approach is to reply to the DATA verb with an accepted status, then apply individual filters and generate a non-delivery report for recipients whose filters rejected the message. But this would be exactly how individual mail filtering is currently implemented, and it carries with it the same disadvantages and setbacks. Another possible approach is to go ahead and issue an extended reply, which will be interpreted as a permanent failure by the sending SMTP Expires XXX XX, XXXX [Page 5] EXDATA SMTP Extension S. Varshavchik XXX XX, XXXX agent. This approach is usually unacceptable because the sending SMTP agent will correctly conclude that every recipient was undeliverable. If a message comes from a mailing list, the mailing list software can end up removing all recipients from the mailing list only because of a malfunctioning mail filter used by one recipient only. Here’s an approach that attempts to achieve a compromise between having recipient-specific mail filtering, and avoiding unwanted disruptions of this kind. A) Recipients must explicitly enable the ability to set filtering rules for their mailboxes B) Allow each recipient to specify "whitelisted" senders and/or network addresses. The mail server will not do any filtering on white-listed mail, and will never send EXDATA-formatted replies to those sources. C) When the first valid RCPT TO: verb is received (after zero or more RCPT TO: verbs that were deemed to be unacceptable for any other reason), determine if the first recipient has this sender whitelisted. D) If the message is whitelisted by the first accepted recipient, any additional recipients MUST also have the message similarly whitelisted. If not, reject the corresponding RCPT TO: verb with a temporary 421 status code. The sending SMTP agent is expected to retransmit the message for all non whitelisted recipients after a small delay. E) If the message is not whitelisted by the first recipient, any additional recipients MUST NOT have this message whitelisted If they do, reject the corresponding RCPT TO: verb with a temporary 421 status code. The sending SMTP agent is expected to retransmit the message for all whitelisted recipients after a small delay. E) This logic results in separate delivery attempts for whitelisted and non whitelisted recipients (with a small delay in between). Individual recipients can decide whether to whitelist specific sources of mail (such as mailing lists). By whitelisting known mail sources, the recipients can make sure that anyone else’s mail filter will not affect their mail delivery. 8.2 Response time to SMTP DATA Expires XXX XX, XXXX [Page 6] EXDATA SMTP Extension S. Varshavchik XXX XX, XXXX Implementations should try to avoid any lengthy delays due to mail filters being used, before returning a reply to the SMTP DATA. Many SMTP clients wait for a response for only a short period of time, before giving up. This can lead to a message duplication problem, described in [RFC1047]. 9. Comments An experimental patch was made available in September 1997 (with various revisions that followed) to a relatively popular mail server. The patch used an XEXDATA EHLO keyword. The experimental patch was mainly used to implement recipient-specific mail filtering: permitting individual recipients of a message to selectively reject incoming E-mail traffic based on the contents of the message. The patch implemented the XEXDATA extension only for the receiving side, not for the sending side. No issues have been reported concerning the theoretical design of the extended result code. 10. References [RFC1425] Klensin, J., Freed, N., Rose, M., Stefferud, E., Crocker, D. "SMTP Service Extensions", RFC 1425, United Nations University, Innosoft International, Inc., Dover Beach Consulting, Inc., Network Management Associates, Inc., The Branch Office, February 1993 [RFC821] Postel, J., "Simple Mail Transfer Protocol", STD 10, RFC 821, USC/Information Sciences Institute, August 1982. [RFC2033] Myers, J., "Local Mail Transfer Protocol", RFC 2033, Carnegie Mellon, October 1996. [RFC1047] Partridge, C., "DUPLICATE MESSAGES AND SMTP", RFC 1047, CIC at BBN Labs, February 1988. 11. Author’s address Sam Varshavchik Double Precision, Inc. 402 Main Street Suite 100-241 Metuchen, NJ 08840 Expires XXX XX, XXXX [Page 7]