]> Kitterman Technical Services

3611 Scheel Dr Ellicott City MD 21042 United States of America scott@kitterman.com

Network Working Group Email on the Internet can be forged in a number of ways. In particular, existing protocols place no restriction on what a sending host can use as the author's address of a message or the domain given on the SMTP HELO/EHLO commands. This document describes version 1 of the Sender Policy Framework (SPF) protocol, whereby a domain can explicitly authorize the hosts that are allowed to use its domain name, and a receiving host can check such authorization. This document obsoletes RFC4408.

The current email infrastructure has the property that any host injecting mail into the system can use any DNS domain name it wants in each of the various identifiers specified by and , plus the "reverse pointer" which is the purported hostname of the SMTP client as far as the DNS is concerned . Although this feature is desirable in some circumstances, it is a major obstacle to reducing Unsolicited Bulk email (UBE, aka spam). Furthermore, many domain owning ADMDs (ADministrative Management Domains, see ) are understandably concerned about the ease with which other entities can make use of their domain names, often with malicious intent. This document defines a protocol by which ADMDs can authorize hosts to use their domain names in the RFC5321.MailFrom or RFC5321.HELO/.EHLO identities. Compliant ADMDs publish Sender Policy Framework (SPF) records in DNS specifying which hosts are permitted to use their names, and compliant mail receivers use the published SPF records to test the authorization of sending Mail Transfer Agents (MTAs) using a given RFC5321.HELO/.EHLO or RFC5321.MailFrom identity during a mail transaction. An additional benefit to mail receivers is that after the use of an identity is verified, local policy decisions about the mail can be made based on the sender's domain, rather than the host's IP address. This is advantageous because reputation of domain names is likely to be more accurate than reputation of host IP addresses. Furthermore, if a claimed identity fails verification, local policy can take stronger action against such email, such as rejecting it.

SPF has been in development since the summer of 2003 and has seen deployment beyond the developers beginning in December 2003. The design of SPF slowly evolved until the spring of 2004 and has since stabilized. There have been quite a number of forms of SPF, some written up as documents, some submitted as Internet Drafts, and many discussed and debated in development forums. The protocol was originally defined in , which this document replaces. The goal of this work is to clearly document the protocol defined by earlier draft specifications of SPF as used in existing implementations. This conception of SPF is sometimes called "SPF Classic". It is understood that particular implementations and deployments will differ from, and build upon, this work. It is hoped that we have nonetheless captured the common understanding of SPF version 1.

This document updates and replaces RFC 4408 that was part of a group of simultaneously published Experimental RFCs (RFC 4405, RFC 4406, RFC 4407, and RFC 4408) in 2006. At that time the IESG requested the community observe the success or failure of the two approaches documented in these RFCs during the two years following publication, in order that a community consensus could be reached in the future. SPF is widely deployed by large and small email providers alike. There are multiple, interoperable implementations. For SPF (as documented in RFC 4408) a careful effort was made to collect and document lessons learned and errata during the two year period. The errata list has been stable (no new submissions) and only minor protocol lessons learned were identified. Resolution of the IESG's experiment is documented in .

The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in .

The ABNF tokens "ALPHA", "DIGIT", and "SP" are defined in . The token "local-part" is defined in . "dot-atom", "quoted-string", "comment", "CFWS", "FWS", and "CRLF" are defined in .

This document is concerned with the portion of a mail message commonly called "envelope sender", "return path", "reverse path", "bounce address", "5321 FROM", or "MAIL FROM". Since these terms are either not well defined or often used casually, this document uses RFC5321.MailFrom as defined in . Note that other terms that might superficially look like the common terms, such as "reverse-path", are used only with the defined meanings from normative documents.

This document also makes use of the HELO/EHLO identity. The "HELO" identity derives from either the SMTP HELO or EHLO command (see ). Since HELO and EHLO can, in many cases, be used interchangeably, they are identified as RFC5321.HELO/.EHLO in . These commands supply the identity of the SMTP client (sending host) for the SMTP session.

There are features that are marked "deprecated". In the context of this document, deprecated means that senders SHOULD NOT publish SPF records that make use of such features because they might be removed entirely in future updates to the protocol. Such features do, however, remain part of the SPF protocol and receiving systems MUST support them unless this document explicitly says otherwise.

It is RECOMMENDED that SPF clients not only check the RFC5321.MailFrom identity, but also separately check the RFC5321.HELO/.EHLO identity by applying the check_host() function () to the RFC5321.HELO/.EHLO identity as the <sender>. Checking RFC5321.HELO/.EHLO promotes consistency of results and can reduce DNS resource usage. Additionally, since SPF records published for RFC5321.HELO/.EHLO identities refer to a single host, when available, they are a very reliable source of host authorization status. Note that requirements for the domain presented in the EHLO or HELO command are not always clear to the sending party, and SPF clients MUST be prepared for the RFC5321.HELO/.EHLO identity to be malformed or an IP address literal.

allows the reverse-path to be null (see Section 4.5.5 in ). In this case, there is no explicit sender mailbox, and such a message can be assumed to be a notification message from the mail system itself. When the reverse-path is null, this document defines the RFC5321.MailFrom identity to be the mailbox composed of the local-part "postmaster" and the RFC5321.HELO/.EHLO identity (which might or might not have been checked separately before). SPF clients MUST check the "RFC5321.MailFrom identity if a completed RFC5321.HELO/.EHLO check has not reached a definitive policy result by applying the check_host() function to the RFC5321.MailFrom identity as the <sender>.

An SPF-compliant domain MUST publish a valid SPF record as described in . This record authorizes the use of the domain name in the RFC5321.HELO/.EHLO and RFC5321.MailFrom identities by the MTAs it specifies. SPF results can be used to make both positive (source is authorized) and negative (source is not authorized) determinations. If domain owners choose to publish SPF records and want to support receivers making negative authorization determinations, then they MUST publish records that end in "-all", or redirect to other records that do, otherwise, no definitive determination of authorization can be made. Potential issues and mitigations associated with negative determinations are discussed in . Domain holders can publish SPF records that explicitly authorize no hosts if mail is not expected to originate using that domain. When changing SPF records, care must be taken to ensure that there is a transition period so that the old policy remains valid until all legitimate email can reasonably expect to have been checked. This can be as much as 30 days.

A mail receiver can perform a set of SPF checks for each mail message it receives. An SPF check tests the authorization of a client host to emit mail with a given identity. Typically, such checks are done by a receiving MTA, but can be performed elsewhere in the mail processing chain so long as the required information is available and reliable. At least the RFC5321.MailFrom identity MUST be checked, but it is RECOMMENDED that the RFC5321.HELO/.EHLO identity also be checked beforehand. Without explicit approval of the domain owner, checking other identities against SPF version 1 records is NOT RECOMMENDED because there are cases that are known to give incorrect results. For example, almost all mailing lists rewrite the RFC5321.MailFrom identity (see ), but some do not change any other identities in the message. The scenario described in , sub-section 1.2, is another example. Documents that define other identities will have to define the method for explicit approval. It is possible that mail receivers will use the SPF check as part of a larger set of tests on incoming mail. The results of other tests might influence whether or not a particular SPF check is performed. For example, finding the sending host's IP address on a local white list might cause all other tests to be skipped and all mail from that host to be accepted. When a mail receiver decides to perform an SPF check, it MUST use a correctly-implemented check_host() function () evaluated with the correct parameters. Although the test as a whole is optional, once it has been decided to perform a test it must be performed as specified so that the correct semantics are preserved between publisher and receiver. To make the test, the mail receiver MUST evaluate the check_host() function with the arguments set as follows: - the IP address of the SMTP client that is emitting the      mail, either IPv4 or IPv6. - the domain portion of the RFC5321.MailFrom or RFC5321.HELO/.EHLO identity. - the RFC5321.MailFrom or RFC5321.HELO/.EHLO identity. Note that the <domain> argument might not be a well-formed domain name. For example, if the reverse-path was null, then the EHLO/HELO domain is used, with its associated problems (see ). In these cases, check_host() is defined in to return a "none" result. Although invalid, malformed, or non-existent domains cause SPF checks to return "none" because no SPF record can be found, it has long been the policy of many MTAs to reject email from such domains, especially in the case of invalid RFC5321.MailFrom. Rejecting email will prevent one method of circumventing of SPF records. Implementations MUST take care to correctly extract the <domain> from the data given with the SMTP MAIL FROM command as many MTAs will still accept such things as source routes (see , Appendix C), the %-hack (see ), and bang paths (see ). These archaic features have been maliciously used to bypass security systems.

This section describes how software that performs the authorization should interpret the results of the check_host() function. The authorization check SHOULD be performed during the processing of the SMTP transaction that sends the mail. This allows errors to be returned directly to the sending MTA by way of SMTP replies. Performing the authorization other than using the return-path and client address at the time of the MAIL command during the SMTP transaction can cause problems, such as the following: (1) It might be difficult to accurately extract the required information from potentially deceptive headers; (2) legitimate email might fail because the sender's policy had since changed. Generating non-delivery notifications to forged identities that have failed the authorization check is a source of backscatter and SHOULD be avoided. section 2 describes backscatter and the problems it causes.

A result of "none" means either (a) no syntactically valid DNS domain name was extracted from the SMTP session that could be used as the one to be authorized, or (b) no TXT records were retrieved from the DNS that appeared to be intended for use by SPF verifiers.

The domain owner has explicitly stated that they cannot or do not want to assert whether the IP address is authorized or not. A "neutral" result MUST be treated exactly like the "none" result; the distinction exists only for informational purposes. Treating "neutral" more harshly than "none" would discourage domain owners from testing the use of SPF records (see ).

A "pass" result means that the client is authorized to inject mail with the given identity. The domain can now, in the sense of reputation, be considered responsible for sending the message. Further policy checks can now proceed with confidence in the legitimate use of the identity.

A "fail" result is an explicit statement that the client is not authorized to use the domain in the given identity. Disposition of SPF fail messages is a matter of local policy. See for considerations on developing local policy. If the checking software chooses to reject the mail during the SMTP transaction, then it SHOULD use an SMTP reply code of 550 (see ) and, if supported, the 5.7.1 enhanced status code (see ), in addition to an appropriate reply text. The check_host() function will return either a default explanation string or one from the domain that published the SPF records (see ). If the information does not originate with the checking software, it should be made clear that the text is provided by the sender's domain. For example:

550-5.7.1 SPF MAIL FROM check failed: 550-5.7.1 The domain example.com explains: 550 5.7.1 Please see http://www.example.com/mailpolicy.html

A "softfail" result should be treated as somewhere between a "fail" and a "neutral". The domain believes the host is not authorized but is not willing to make a strong policy statement. Receiving software SHOULD NOT reject the message based solely on this result, but MAY subject the message to closer scrutiny than normal. The domain owner wants to discourage the use of this host and thus desires limited feedback when a "softfail" result occurs. For example, the recipient's Mail User Agent (MUA) could highlight the "softfail" status, or the receiving MTA could give the sender a message using greylisting, , with a note the first time the message is received, but accept it on a later attempt based on receiver policy.

A "temperror" result means the SPF client encountered a transient (DNS) error while performing the check. Checking software can choose to accept or temporarily reject the message. If the message is rejected during the SMTP transaction for this reason, the software SHOULD use an SMTP reply code of 451 and, if supported, the 4.4.3 enhanced status code. These errors can be caused by DNS problems in either the sender's or receiver's DNS.

A "permerror" result means the domain's published records could not be correctly interpreted. This signals an error condition that definitely requires manual intervention to be resolved. If the message is rejected during the SMTP transaction for this reason, the software SHOULD use an SMTP reply code of 550 and, if supported, the 5.5.2 enhanced status code. Be aware that if the domain owner uses macros (), it is possible that this result is due to the checked identities having an unexpected format.

An SPF record is a DNS TXT (type 16) Resource Record (RR) that declares which hosts are, and are not, authorized to use a domain name for the RFC5321.HELO/.EHLO and RFC5321.MailFrom identities. Loosely, the record partitions all hosts into permitted and not-permitted sets (though some hosts might fall into neither category). The SPF record is a single string of text. An example record is the following: v=spf1 +mx a:colo.example.com/28 -all This record has a version of "spf1" and three directives: "+mx", "a:colo.example.com/28" (the + is implied), and "-all". Domain owners wishing to be SPF compliant MUST publish SPF records for the hosts that are used in the RFC5321.MailFrom and RFC5321.HELO/.EHLO identities. The SPF records are placed in the DNS tree at the host name it pertains to, not a subdomain under it, such as is done with SRV records . The example in might be published via these lines in a domain zone file:

example.com. TXT "v=spf1 +mx a:colo.example.com/28 -all" smtp-out.example.com. TXT "v=spf1 a -all"

Since TXT records have multiple uses, beware of other TXT records published there for other purposes. They might cause problems with size limits (see ) and care MUST be taken to ensure only SPF records are used for SPF processing. Domains publishing records SHOULD try to keep the number of "include" mechanisms and chained "redirect" modifiers to a minimum. Domains SHOULD also try to minimize the amount of other DNS information needed to evaluate a record. provides some suggestions on how to achieve this.

SPF records MUST be published as type TXT . The character content of the record is encoded as . Use of alternate DNS RR types was supported in SPF's experimental phase, but has been discontinued. See Appendix A of for further information.

A domain name MUST NOT have multiple records that would cause an authorization check to select more than one record. See for the selection rules.

As defined in sections 3.3.14 and 3.3, a single text DNS record (either TXT or SPF RR types) can be composed of more than one string. If a published record contains multiple character-strings, then the record MUST be treated as if those strings are concatenated together without adding spaces. For example: IN TXT "v=spf1 .... first" "second string..." MUST be treated as equivalent to IN TXT "v=spf1 .... firstsecond string..." TXT records containing multiple strings are useful in constructing records that would exceed the 255-byte maximum length of a character-string within a single TXT record.

The published SPF record for a given domain name SHOULD remain small enough that the results of a query for it will fit within 512 octets. This UDP limit is defined in section 2.3.4. This will keep even older DNS implementations from falling over to TCP. Since the answer size is dependent on many things outside the scope of this document, it is only possible to give this guideline: If the combined length of the DNS name and the text of all the records of a given type is under 450 characters, then DNS answers ought to fit in UDP packets. Note that when computing the sizes for queries of the TXT format, one must take into account any other TXT records published at the domain name. Records that are too long to fit in a single UDP packet could be silently ignored by SPF clients due to firewall and other issues that cause DNS over TCP to be less reliable than DNS over UDP.

Use of wildcard records for publishing is NOT RECOMMENDED. Care must be taken if wildcard records are used. If a domain publishes wildcard MX records, it might want to publish wildcard declarations, subject to the same requirements and problems. In particular, the declaration MUST be repeated for any host that has any RR records at all, and for subdomains thereof. For example, the example given in , Section 4.3.3, could be extended with the following:

X.COM. MX 10 A.X.COM X.COM. TXT "v=spf1 a:A.X.COM -all" *.X.COM. MX 10 A.X.COM *.X.COM. TXT "v=spf1 a:A.X.COM -all" A.X.COM. A 1.2.3.4 A.X.COM. MX 10 A.X.COM A.X.COM. TXT "v=spf1 a:A.X.COM -all" *.A.X.COM. MX 10 A.X.COM *.A.X.COM. TXT "v=spf1 a:A.X.COM -all"

Notice that SPF records MUST be repeated twice for every name within the domain: once for the name, and once with a wildcard to cover the tree under the name. Use of wildcards is discouraged in general as they cause every name under the domain to exist and queries against arbitrary names will never return RCODE 3 (Name Error) and will interfere with existence tests such as the one described in .

This description is not an API (Application Program Interface) definition, but rather a function description used to illustrate the algorithm. A compliant SPF implementation MUST do something semantically equivalent to this description. The check_host() function fetches SPF records, parses them, and evaluates them to determine whether a particular host is or is not permitted to send mail with a given identity. Mail receivers that perform this check MUST correctly evaluate the check_host() function as described here. Implementations MAY use a different algorithm than the canonical algorithm defined here, so long as the results are the same in all cases.

The check_host() function takes these arguments: - the IP address of the SMTP client that is emitting the      mail, either IPv4 or IPv6. - the domain that provides the sought-after authorization     information; initially, the domain portion of the "MAIL     FROM" or RFC5321.HELO/.EHLO identity. - the RFC5321.MailFrom or RFC5321.HELO/.EHLO identity. The domain portion of <sender> will usually be the same as the <domain> argument when check_host() is initially evaluated. However, this will generally not be true for recursive evaluations (see below). Actual implementations of the check_host() function might need additional arguments.

The function check_host() can return one of several results described in . Based on the result, the action to be taken is determined by the local policies of the receiver.

If the <domain> is malformed (e.g. label longer than 63 characters, zero-length label not at the end, etc.) or is not a fully qualified domain name, or if the DNS lookup returns "domain does not exist" (RCODE 3), check_host() immediately returns the result "none". Properly formed domains are fully qualified email domains as described in Section 2.3.5. Internationalized domain names MUST be encoded as A-labels, as described in Section 2.3 of . If the <sender> has no local-part, substitute the string "postmaster" for the local-part.

In accordance with how the records are published (see above), a DNS query needs to be made for the <domain> name, querying for type TXT only. If all DNS lookups that are made return a server failure (RCODE 2), or other error (RCODE other than 0 or 3), or time out, then check_host() terminates immediately with the result "temperror". Alternatively, for a server failure (RCODE 2) result, check_host() MAY track failures and treat multiple failures within 24 hours for the same domain as "permerror". This alternative is intended to shorten the queue time of messages that cannot be accepted, by returning a permanent negative completion reply code to the client, instead of a transient one. suggests on an algorithm for doing such tracking and handling of server failure codes.

Records begin with a version section:

Starting with the set of records that were returned by the lookup, discard records that do not begin with a version section of exactly "v=spf1". Note that the version section is terminated either by an SP character or the end of the record. A record with a version section of "v=spf10" does not match and must be discarded. If the resultant record set includes no records, check_host() produces the "none" result. If the resultant record set includes more than one record, check_host() produces the "permerror" result.

After one SPF record has been selected, the check_host() function parses and interprets it to find a result for the current test. If there are any syntax errors, check_host() returns immediately with the result "permerror". Implementations MAY choose to parse the entire record first and return "permerror" if the record is not syntactically well formed. However, in all cases, any syntax errors anywhere in the record MUST be detected.

There are two types of terms: mechanisms and modifiers. A record contains an ordered list of these as specified in the following Augmented Backus-Naur Form (ABNF).

Most mechanisms allow a ":" or "/" character after the name. Modifiers always contain an equals ('=') character immediately after the name, and before any ":" or "/" characters that might be part of the macro-string. Terms that do not contain any of "=", ":", or "/" are mechanisms, as defined in . As per the definition of the ABNF notation in , mechanism and modifier names are case-insensitive.

Each mechanism is considered in turn from left to right. If there are no more mechanisms, the result is specified in . When a mechanism is evaluated, one of three things can happen: it can match, not match, or throw an exception. If it matches, processing ends and the qualifier value is returned as the result of that record. If it does not match, processing continues with the next mechanism. If it throws an exception, mechanism processing ends and the exception value is returned. The possible qualifiers, and the results they cause check_host() to return are as follows:

"+" pass "-" fail "~" softfail "?" neutral

The qualifier is optional and defaults to "+". When a mechanism matches and the qualifier is "-", then a "fail" result is returned and the explanation string is computed as described in . The specific mechanisms are described in .

Modifiers are not mechanisms. They do not return match or not-match. Instead, they provide additional information. Although modifiers do not directly affect the evaluation of the record, the "redirect" modifier has an effect after all the mechanisms have been evaluated.

SPF implementations MUST limit the number of mechanisms and modifiers ("terms") that DNS lookups to at most 10 during SPF evaluation. Specifically, the "include", "a", "mx", "ptr", and "exists" mechanisms as well as the "redirect" modifier count against this limit. The "all", "ip4", and "ip6" mechanisms do not count against this limit. If this number is exceeded during a check, a permerror MUST be returned. The "exp" modifier does not count against this limit because the DNS lookup to fetch the explanation string occurs after the SPF record evaluation has been completed. When evaluating the "mx" and "ptr" mechanisms, or the %{p} macro, there MUST be a limit of no more than 10 MX or PTR RRs looked up and checked. If more than 10 "mx" or "ptr" records are returned for this further lookup, a permerror MUST be returned. This limit is per mechanism or macro in the record and in addition to the lookup limits above.

If none of the mechanisms match and there is no "redirect" modifier, then the check_host() returns a result of "neutral", just as if "?all" were specified as the last directive. If there is a "redirect" modifier, check_host() proceeds as defined in . Note that records SHOULD always use either a "redirect" modifier or an "all" mechanism to explicitly terminate processing. Although the latter has default (specifically "?all"), it aids debugging efforts if it is explicitly included. For example: v=spf1 +mx -all or v=spf1 +mx redirect=_spf.example.com

Several of these mechanisms and modifiers have a domain-spec section. The domain-spec string is subject to macro expansion (see ). The resulting string is the common presentation form of a fully-qualified DNS name: a series of labels separated by periods. This domain is called the <target-name> in the rest of this document. Note: The result of the macro expansion is not subject to any further escaping. Hence, this facility cannot produce all characters that are legal in a DNS label (e.g., the control characters). However, this facility is powerful enough to express legal host names and common utility labels (such as "_spf") that are used in DNS. For several mechanisms, the <domain-spec> is optional. If it is not provided, the <domain> is used as the <target-name>. Domain and domain-spec are syntactically identical after macro expansion. Domain is an input value for check_host() while domain-spec is computed by check_host().

This section defines two types of mechanisms. Basic mechanisms contribute to the language framework. They do not specify a particular type of authorization scheme.

all include

Designated sender mechanisms are used to designate a set of <ip> addresses as being permitted or not permitted to use the <domain> for sending mail.

a mx ptr (deprecated) ip4 ip6 exists

The following conventions apply to all mechanisms that perform a comparison between <ip> and an IP address at any point: If no CIDR prefix length is given in the directive, then <ip> and the IP address are compared for equality. (Here, CIDR is Classless Inter-Domain Routing, described in .) If a CIDR prefix length is specified, then only the specified number of high-order bits of <ip> and the IP address are compared for equality. When any mechanism fetches host addresses to compare with <ip>, when <ip> is an IPv4 address, A records are fetched; when <ip> is an IPv6 address, AAAA records are fetched. Even if the SMTP connection uses IPv6, an IPv4-mapped IPv6 IP address (see , Section 2.5.5) MUST still be considered an IPv4 address and MUST be evaluated using IPv4 mechanisms (i.e. "ip4" and "a"). Several mechanisms rely on information fetched from the DNS. For these DNS queries, except where noted, if the DNS server returns an error (RCODE other than 0 or 3) or the query times out, the mechanism throws the exception "temperror". If the server returns "domain does not exist" (RCODE 3), then evaluation of the mechanism continues as if the server returned no error (RCODE 0) and zero answer records.

The "all" mechanism is a test that always matches. It is used as the rightmost mechanism in a record to provide an explicit default. For example: v=spf1 a mx -all Mechanisms after "all" will never be tested. Mechanisms listed after "all" MUST be ignored. Any "redirect" modifier () has MUST be ignored when there is an "all" mechanism in the record.

The "include" mechanism triggers a recursive evaluation of check_host(). The domain-spec is expanded as per . Check_host() is evaluated with the resulting string as the <domain>. The <ip> and <sender> arguments remain the same as in the current evaluation of check_host(). The recursive evaluation returns either match, not match, or an error. If it matches, then the appropriate result for the include: mechanism is used (e.g. include or +include gives a "pass" result and -include gives "fail). If there is no match, the parent check_host() resumes processing as per the table below, with the previous value of <domain> restored. In hindsight, the name "include" was poorly chosen. Only the evaluated result of the referenced SPF record is used, rather than acting as if the referenced SPF record was literally included in the first. For example, evaluating a "-all" directive in the referenced record does not terminate the overall processing and does not necessarily result in an overall "fail". (Better names for this mechanism would have been "if-match", "on-match", etc.) The "include" mechanism makes it possible for one domain to designate multiple administratively-independent domains. For example, a vanity domain "example.net" might send mail using the servers of administratively-independent domains example.com and example.org. Example.net could say IN TXT "v=spf1 include:example.com include:example.org -all" This would direct check_host() to, in effect, check the records of example.com and example.org for a "pass" result. Only if the host were not permitted for either of those domains would the result be "fail". Whether this mechanism matches, does not match, or throws an exception depends on the result of the recursive evaluation of check_host(): A recursive check_host() result of: Causes the "include" mechanism to: pass match fail not match softfail not match neutral not match temperror throw temperror permerror throw permerror none throw permerror The "include" mechanism is intended for crossing administrative boundaries. For example, if example.com and example.org were managed by the same entity, and if the permitted set of hosts for both domains was "mx:example.com", it would be possible for example.org to specify "include:example.com", but it would be preferable to specify "redirect=example.com" or even "mx:example.com". With the "include" mechanism an administratively external set of hosts can be authorized, but determination of sender policy is still a function of the original domain's SPF record (as determined by the "all" mechanism in that record). The redirect modifier is more suitable for consolidating both authorizations and policy into a common set to be shared within an ADMD. Redirect is much more like a common code element to be shared among records in a single ADMD. It is possible to control both authorized hosts and policy for an arbitrary number of domains from a single record.

This mechanism matches if <ip> is one of the <target-name>'s IP addresses.

An address lookup is done on the <target-name>. The <ip> is compared to the returned address(es). If any address matches, the mechanism matches.

This mechanism matches if <ip> is one of the MX hosts for a domain name.

check_host() first performs an MX lookup on the <target-name>. Then it performs an address lookup on each MX name returned. The <ip> is compared to each returned IP address. To prevent Denial of Service (DoS) attacks, more than 10 MX names MUST NOT be looked up during the evaluation of an "mx" mechanism (see ). If any address matches, the mechanism matches. Note regarding implicit MXs: If the <target-name> has no MX records, check_host() MUST NOT pretend the target is its single MX, and MUST NOT default to an A or AAAA lookup on the <target-name> directly. This behavior diverges from the legacy "implicit MX" rule, (See , Section 5. If such behavior is desired, the publisher should specify an "a" directive).

This mechanism tests whether the DNS reverse-mapping for <ip> exists and correctly points to a domain name within a particular domain. This mechanism is deprecated and SHOULD NOT be used.

The <ip>'s name is looked up using this procedure: Perform a DNS reverse-mapping for <ip> Look up the corresponding PTR record in "in-addr.arpa." if the address is an IPv4 one and in "ip6.arpa." if it is an IPv6 address. For each record returned, validate the domain name by looking up its IP address. To prevent DoS attacks, more than 10 PTR names MUST NOT be looked up during the evaluation of a "ptr" mechanism (see ). If <ip> is among the returned IP addresses, then that domain name is validated. Check all validated domain names to see if they either match the <target-name> domain or are a subdomain of the <target-name> domain. If any do, this mechanism matches. If no validated domain name can be found, or if none of the validated domain names match or are a subdomain of the <target-name>, this mechanism fails to match. If a DNS error occurs while doing the PTR RR lookup, then this mechanism fails to match. If a DNS error occurs while doing an A RR lookup, then that domain name is skipped and the search continues.

Pseudocode: sending-domain_names := ptr_lookup(sending-host_IP); if more than 10 sending-domain_names are found, use at most 10. for each name in (sending-domain_names) { IP_addresses := a_lookup(name); if the sending-domain_IP is one of the IP_addresses { validated-sending-domain_names += name; } } for each name in (validated-sending-domain_names) { if name ends in <domain-spec>, return match. if name is <domain-spec>, return match. } return no-match.

This mechanism matches if the <target-name> is either a subdomain of a validated domain name or if the <target-name> and a validated domain name are the same. For example: "mail.example.com" is within the domain "example.com", but "mail.bad-example.com" is not. Note: This mechanism has been deprecated because it is slow, it is not as reliable as other mechanisms in cases of DNS errors, and it places a large burden on the .arpa name servers. If used, proper PTR records must be in place for the domain's hosts and the "ptr" mechanism should be one of the last mechanisms checked. After many yearrs of SPF deployment experience it has been concluded it is unnecessary and more reliable alternatives used instead. It is, however, still in use and part of the SPF protocol, so compliant check_host() implementations MUST support it.

These mechanisms test whether <ip> is contained within a given IP network.

; e.g., 2001:DB8::CD30 ]]>

The <ip> is compared to the given network. If CIDR prefix length high-order bits match, the mechanism matches. If ip4-cidr-length is omitted, it is taken to be "/32". If ip6-cidr-length is omitted, it is taken to be "/128". It is not permitted to omit parts of the IP address instead of using CIDR notations. That is, use 192.0.2.0/24 instead of 192.0.2.

This mechanism is used to construct an arbitrary domain name that is used for a DNS A record query. It allows for complicated schemes involving arbitrary parts of the mail envelope to determine what is permitted.

The domain-spec is expanded as per . The resulting domain name is used for a DNS A RR lookup. If any A record is returned, this mechanism matches. The lookup type is A even when the connection type is IPv6. Domains can use this mechanism to specify arbitrarily complex queries. For example, suppose example.com publishes the record: v=spf1 exists:%{ir}.%{l1r+-}._spf.%{d} -all The <target-name> might expand to "1.2.0.192.someuser._spf.example.com". This makes fine-grained decisions possible at the level of the user and client IP address. This mechanism enables queries that mimic the style of tests that existing DNS white/black lists (DNSxLs) use, as described in . The query will either return NXDOMAIN (no match), any valid answer (match), or an error.

Modifiers are name/value pairs that provide additional information. Modifiers always have an "=" separating the name and the value. The modifiers defined in this document ("redirect" and "exp") MAY appear anywhere in the record, but SHOULD appear at the end, after all mechanisms. Ordering of these two modifiers does not matter. These two modifiers MUST NOT appear in a record more than once each. If they do, then check_host() exits with a result of "permerror". Unrecognized modifiers MUST be ignored no matter where in a record, or how often. This allows implementations of this document to gracefully handle records with modifiers that are defined in other specifications.

The redirect modifier intended for consolidating both authorizations and policy into a common set to be shared within a single ADMD. Redirect is like a common code element to be shared among records in a single ADMD. It is possible to control both authorized hosts and policy for an arbitrary number of domains from a single record.

If all mechanisms fail to match, and a "redirect" modifier is present, then processing proceeds as follows: The domain-spec portion of the redirect section is expanded as per the macro rules in . Then check_host() is evaluated with the resulting string as the <domain>. The <ip> and <sender> arguments remain the same as in the current evaluation of check_host(). The result of this new evaluation of check_host() is then considered the result of the current evaluation with the exception that if no SPF record is found, or if the target-name is malformed, the result is a "permerror" rather than "none". Note that the newly-queried domain can itself specify redirect processing. This facility is intended for use by organizations that wish to apply the same record to multiple domains. For example:

la.example.com. TXT "v=spf1 redirect=_spf.example.com" ny.example.com. TXT "v=spf1 redirect=_spf.example.com" sf.example.com. TXT "v=spf1 redirect=_spf.example.com" _spf.example.com. TXT "v=spf1 mx:example.com -all"

In this example, mail from any of the three domains is described by the same record. This can be an administrative advantage. Note: In general, the domain "A" cannot reliably use a redirect to another domain "B" not under the same administrative control. Since the <sender> stays the same, there is no guarantee that the record at domain "B" will correctly work for mailboxes in domain "A", especially if domain "B" uses mechanisms involving local-parts. An "include" directive is generally be more appropriate. For clarity, it is RECOMMENDED that any "redirect" modifier appear as the very last term in a record.

If check_host() results in a "fail" due to a mechanism match (such as "-all"), and the "exp" modifier is present, then the explanation string returned is computed as described below. If no "exp" modifier is present, then either a default explanation string or an empty explanation string MUST be returned. The domain-spec is macro expanded (see ) and becomes the <target-name>. The DNS TXT record for the <target-name> is fetched. If there are any DNS processing errors (any RCODE other than 0), or if no records are returned, or if more than one record is returned, or if there are syntax errors in the explanation string, then proceed as if no exp modifier was given. The fetched TXT record's strings are concatenated with no spaces, and then treated as an explain-string, which is macro-expanded. This final result is the explanation string. Implementations MAY limit the length of the resulting explanation string to allow for other protocol constraints and/or reasonable processing limits. Since the explanation string is intended for an SMTP response and Section 2.4 says that responses are in , the explanation string MUST be limited to US-ASCII. Software evaluating check_host() can use this string to communicate information from the publishing domain in the form of a short message or URL. Software SHOULD make it clear that the explanation string comes from a third party. For example, it can prepend the macro string "%{o} explains: " to the explanation, such as shown in . Suppose example.com has this record: v=spf1 mx -all exp=explain._spf.%{d} Here are some examples of possible explanation TXT records at explain._spf.example.com: "Mail from example.com should only be sent by its own servers." a simple, constant message "%{i} is not one of %{d}'s designated mail servers." a message with a little more information, including the IP address that failed the check "See http://%{d}/why.html?s=%{S}&i=%{I}" a complicated example that constructs a URL with the arguments to check_host() so that a web page can be generated with detailed, custom instructions Note: During recursion into an "include" mechanism, an exp= modifier from the <target-name> MUST NOT be used. In contrast, when executing a "redirect" modifier, an exp= modifier from the original domain MUST NOT be used.

It is RECOMMENDED that SMTP receivers record the result of SPF processing in the message header. There are two methods for doing this: the Received-SPF Header Field defined here and the more generic Authentication-Results Header Field defined in . Because these fields are generally used within a receiving ADMD, it is a local policy choice which to include. In general, the more broadly applicable Authentication-Results Header Field ought to be used, but it SHOULD provide the same information as if a Received-SPF Header Field were used. If an SMTP receiver chooses to do so, it SHOULD use one of these header fields for each identity that was checked. This information is intended for the recipient. (Information intended for the sender is described in , Explanation.)

The Received-SPF header field is a trace field (see Section 3.6.7) and SHOULD be prepended to the existing header, above the Received: field that is generated by the SMTP receiver. It MUST appear above all other Received-SPF fields in the message. The header field has the following format:

quoted-string = comment = CFWS = FWS = CRLF = ]]>

The header field SHOULD include a "(...)" style comment after the result, conveying supporting information for the result, such as <ip>, <sender>, and <domain>. The following key-value pairs are designed for later machine parsing. SPF clients SHOULD give enough information so that the SPF results can be verified. That is, at least "client-ip", "helo", and, if the RFC5321.MailFrom identity was checked, "envelope-from". the IP address of the SMTP client the envelope sender mailbox the host name given in the HELO or EHLO command the mechanism that matched (if no mechanisms matched, substitute the word "default") if an error was returned, details about the error the host name of the SPF client the identity that was checked; see the <identity> ABNF rule Other keys MAY be defined by SPF clients. SPF clients MUST make sure that the Received-SPF header field does not contain invalid characters, is not excessively long (See Section 2.1.1), and does not contain malicious data that has been provided by the sender. Examples of various header field styles that could be generated are the following:

Received-SPF: pass (mybox.example.org: domain of myname@example.com designates 192.0.2.1 as permitted sender) receiver=mybox.example.org; client-ip=192.0.2.1; envelope-from="myname@example.com"; helo=foo.example.com; Received-SPF: fail (mybox.example.org: domain of myname@example.com does not designate 192.0.2.1 as permitted sender) identity=mailfrom; client-ip=192.0.2.1; envelope-from="myname@example.com";

As mentioned in , the Authentication-Results header field is designed to communicate lists of tests a border MTA did and their results. The specified elements of the field provide less information than SPF-Received:

Authentication-Results: myhost.example.org; spf=pass smtp.mailfrom=example.net Received-SPF: pass (myhost.example.org: domain of myname@example.com designates 192.0.2.1 as permitted sender) receiver=mybox.example.org; client-ip=192.0.2.1; envelope-from="myname@example.com"; helo=foo.example.com;

It is, however, possible to add CFWS in an Authentication-Results Header Field to provide the equivalent information. Receivers SHOULD include the explanation ("myhost.example.org: domain of myname@example.com designates 192.0.2.1 as permitted sender" in the example above), the recipient address ("receiver=mybox.example.org"), the IP address of the sending SMTP client ("client-ip=192.0.2.1"), and both the 5321.MailFrom address and/or the 5321.HELO/.EHLO. Authserv-id SHOULD be the name of the receiving host performing the check. reason SHOULD include a key-value-list with keys provinding information normally included in a Received-SPF Header Field that is not already part of the Authentication-Results Header Field. That is, at least "client-ip", "helo", and, if the RFC5321.MailFrom identity was checked, "envelope-from".

reasonspec = ]]>

A suitably enhanced Authentication-Results Header Field might look like:

Authentication-Results: myhost.example.org; spf=pass reason="client-ip=192.0.2.1" smtp.helo=foo.example.com smtp.mailfrom=user@example.net

Many mechanisms and modifiers perform macro expansion on a term.

A literal "%" is expressed by "%%". "%_" expands to a single " " space. "%-" expands to a URL-encoded space, viz., "%20". The following macro letters are expanded in term arguments: s = <sender> l = local-part of <sender> o = domain of <sender> d = <domain> i = <ip> p = the validated domain name of <ip> (deprecated) v = the string "in-addr" if <ip> is ipv4, or "ip6" if <ip> is ipv6 h = HELO/EHLO domain The following macro letters are allowed only in "exp" text: c = SMTP client IP (easily readable format) r = domain name of host performing the check t = current timestamp A '%' character not followed by a '{', '%', '-', or '_' character is a syntax error. So -exists:%(ir).sbl.spamhaus.example.org is incorrect and will cause check_host() to yield a "permerror". Instead, say -exists:%{ir}.sbl.spamhaus.example.org Optional transformers are the following: = zero or more digits = reverse value, splitting on dots by default If transformers or delimiters are provided, the replacement value for a macro letter is split into parts. After performing any reversal operation and/or removal of left-hand parts, the parts are rejoined using "." and not the original splitting characters. By default, strings are split on "." (dots). Note that no special treatment is given to leading, trailing, or consecutive delimiters in input strings, and so the list of parts might contain empty strings. Some older implementations of SPF prohibit trailing dots in domain names, so trailing dots SHOULD NOT be published by domain owners, although they MUST be accepted by implementations conforming to this document. Macros MAY specify delimiter characters that are used instead of ".". The 'r' transformer indicates a reversal operation: if the client IP address were 192.0.2.1, the macro %{i} would expand to "192.0.2.1" and the macro %{ir} would expand to "1.2.0.192". The DIGIT transformer indicates the number of right-hand parts to use, after optional reversal. If a DIGIT is specified, the value MUST be nonzero. If no DIGITs are specified, or if the value specifies more parts than are available, all the available parts are used. If the DIGIT was 5, and only 3 parts were available, the macro interpreter would pretend the DIGIT was 3. Implementations MUST support at least a value of 128, as that is the maximum number of labels in a domain name. The "s" macro expands to the <sender> argument. It is an email address with a local-part, an "@" character, and a domain. The "l" macro expands to just the local-part. The "o" macro expands to just the domain part. Note that these values remain the same during recursive and chained evaluations due to "include" and/or "redirect". Note also that if the original <sender> had no local-part, the local-part was set to "postmaster" in initial processing (see ). local-parts MUST either be ASCII or encoded as A-labels per Section 2.3 of to be used in the "s" or "l" macros. For IPv4 addresses, both the "i" and "c" macros expand to the standard dotted-quad format. For IPv6 addresses, the "i" macro expands to a dot-format address; it is intended for use in %{ir}. The "c" macro MAY expand to any of the hexadecimal colon-format addresses specified in , Section 2.2. It is intended for humans to read. The "p" macro expands to the validated domain name of <ip>. The procedure for finding the validated domain name is defined in . If the <domain> is present in the list of validated domains, it SHOULD be used. Otherwise, if a subdomain of the <domain> is present, it SHOULD be used. Otherwise, any name from the list MAY be used. If there are no validated domain names or if a DNS error occurs, the string "unknown" is used. This macro is deprecated and SHOULD NOT be used. The "r" macro expands to the name of the receiving MTA. This SHOULD be a fully qualified domain name, but if one does not exist (as when the checking is done by a MUA) or if policy restrictions dictate otherwise, the word "unknown" SHOULD be substituted. The domain name can be different from the name found in the MX record that the client MTA used to locate the receiving MTA. The "t" macro expands to the decimal representation of the approximate number of seconds since the Epoch (Midnight, January 1, 1970, UTC) at the time of the evaluation. This is the same value as is returned by the POSIX time() function in most standards-compliant libraries. When the result of macro expansion is used in a domain name query, if the expanded domain name exceeds 253 characters (the maximum length of a domain name), the left side is truncated to fit, by removing successive domain labels (and their following dots) until the total length does not exceed 253 characters. Uppercased macros expand exactly as their lowercased equivalents, and are then URL escaped. URL escaping must be performed for characters not in the "unreserved" set, which is defined in . Note: Care must be taken so that macro expansion for legitimate email does not exceed the 63-character limit on DNS labels. The local-part of email addresses, in particular, can have more than 63 characters between dots. Note: Domains SHOULD avoid using the "s", "l", "o", or "h" macros in conjunction with any mechanism directive. Although these macros are powerful and allow per-user records to be published, they severely limit the ability of implementations to cache results of check_host() and they reduce the effectiveness of DNS caches. Implementations SHOULD be aware that if no directive processed during the evaluation of check_host() contains an "s", "l", "o", or "h" macro, then the results of the evaluation can be cached on the basis of <domain> and <ip> alone for as long as the shortest Time To Live (TTL) of all the DNS records involved. The macro ABNF does not in itself limit macro expansion so that it results in a valid domain-spec. Nonsense such as this is permitted:

%{s.-+,/_=.+,..+.}

Just because the macro ABNF allows it, does not make it valid. The result of macro expansion MUST still be a valid domain-spec.

The <sender> is strong-bad@email.example.com. The IPv4 SMTP client IP is 192.0.2.3. The IPv6 SMTP client IP is 2001:DB8::CB01. The PTR domain name of the client IP is mx.example.org.

macro expansion ------- ---------------------------- %{s} strong-bad@email.example.com %{o} email.example.com %{d} email.example.com %{d4} email.example.com %{d3} email.example.com %{d2} example.com %{d1} com %{dr} com.example.email %{d2r} example.email %{l} strong-bad %{l-} strong.bad %{lr} strong-bad %{lr-} bad.strong %{l1r-} strong macro-string expansion -------------------------------------------------------------------- %{ir}.%{v}._spf.%{d2} 3.2.0.192.in-addr._spf.example.com %{lr-}.lp._spf.%{d2} bad.strong.lp._spf.example.com %{lr-}.lp.%{ir}.%{v}._spf.%{d2} bad.strong.lp.3.2.0.192.in-addr._spf.example.com %{ir}.%{v}.%{l1r-}.lp._spf.%{d2} 3.2.0.192.in-addr.strong.lp._spf.example.com %{d2}.trusted-domains.example.net example.com.trusted-domains.example.net IPv6: %{ir}.%{v}._spf.%{d2} 1.0.B.C.0.0.0.0. 0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.8.B.D.0.1.0.0.2.ip6._spf.example.com

This section outlines the major implications that adoption of this document will have on various entities involved in Internet email. It is intended to make clear to the reader where this document knowingly affects the operation of such entities. This section is not a "how-to" manual, or a "best practices" document, and it is not a comprehensive list of what such entities should do in light of this document. This section is non-normative. describes the Internet email architecture. This section is organized based on the different segments of the architecture.

Originating ADMDs (ADministrative Management Domains - Section 2.2.1 and Section 2.3) that wish to be compliant with this specification will need to determine the list of relays ( Section 2.2.2) that they allow to use their domain name in the RFC5321.HELO/.EHLO and RFC5321.MailFrom identities when relaying to other ADMDs. It is recognized that forming such a list is not just a simple technical exercise, but involves policy decisions with both technical and administrative considerations.

Minimizing the DNS resources required for SPF lookups can be done by choosing directives that require less DNS information and by placing lower-cost mechanisms earlier in the SPF record.

specifies the limits receivers have to use. It is essential to publish records that do not exceed these requirements. It is also required to carefully weight the cost and the maintainability of licit solutions. For example, consider a domain set up as follows:

example.com. IN MX 10 mx.example.com. IN MX 20 mx2.example.com. mx.example.com. IN A 192.0.2.1 mx2.example.com. IN A 192.0.2.129

Assume the administrative point is to authorize (pass) mx and mx2 while failing every other host. Compare the following solutions:

Best record: example.com. IN TXT "v=spf1 ip4:192.0.2.1 ip4:192.0.2.129 -all"

Good record: $ORIGIN example.com. @ IN TXT "v=spf1 a:authorized_spf.example.com -all" authorized_spf IN A 192.0.2.1 IN A 192.0.2.129

Expensive record: example.com. IN TXT "v=spf1 mx:example.com -all"

Wasteful, bad record: example.com. IN TXT "v=spf1 ip4:192.0.2.0/24 mx -all"

There might be administrative considerations: using "a" over "ip4" or "ip6" allows hosts to be renumbered easily. Using "mx" over "a" allows the set of mail hosts to be changed easily. Unless such changes are common, it is better to use the less resource intensive mechanisms like "ip4" and "ip6". Validating correct deployment is difficult. describes one mechanism for soliciting feedback on SPF failures. Another approach that can be helpful to publish records that include a "tracking exists:" mechanism. By looking at the name server logs, a rough list can then be generated. For example: v=spf1 exists:_h.%{h}._l.%{l}._o.%{o}._i.%{i}._spf.%{d} ?all Regardless of the method used, understanding the ADMD's outbound mail architecture is essential to effective deployment.

As explained in , allows the reverse-path to be null, which is typical of some Delivery Status Notification , commonly called email bounces. In this case the only entity available for performing an SPF check is the RFC5321.HELO/.EHLO identity defined in . SPF functionality is enhanced by administrators ensuring this identity is set correctly and has an appropriate SPF record. It is common to have the HELO identity set to hostname instead of domain. Zone file generation for significant numbers of hosts can be consolidated using the redirect modifier and scripted for initial deployment. Administrators might alternatively consider publishing an SPF for *.domain (wildcard domains) in that case if the involved hostnames do not have already any other DNS record.

Broadly speaking, there are two types of mediating ADMDs that can affect SPF deployment of other ADMDs: mailing lists (see Section 5.3) and ReSenders ( Section 5.2).

Mailing lists must be aware of how they re-inject mail that is sent to the list. Mailing lists MUST comply with the requirements in , Section 3.10, and , Section 5.3.6, that say that the reverse-path MUST be changed to be the mailbox of a person or other entity who administers the list. Whereas the reasons for changing the reverse-path are many and long-standing, SPF adds enforcement to this requirement. In practice, almost all mailing list software in use already complies with this requirement. Mailing lists that do not comply might encounter problems depending on how access to the list is restricted. Such lists that are entirely internal to a domain (only people in the domain can send to or receive from the list) are not affected.

Forwarding services take mail that is received at a mailbox and direct it to some external mailbox. At the time of this writing, the near-universal practice of such services is to use the original RFC5321.MailFrom of a message when re-injecting it for delivery to the external mailbox. and describe this action as an "alias" rather than a "mail list". This means the external mailbox's MTA sees all such mail in a connection from a host of the forwarding service, and so the RFC5321.MailFrom identity will not, in general, pass authorization. There are three places that techniques can be used to ameliorate this problem. The beginning, when email is first sent (Originating ADMDs). "Neutral" results could be given for IP addresses that might be forwarders, instead of "fail" results. For example: "v=spf1 mx -exists:%{ir}.sbl.spamhaus.example.org ?all" This would cause a lookup on an anti-spam DNS blacklist (DNSBL) and cause a result of "fail" only for email coming from listed sources. All other email, including email sent through forwarders, would receive a "neutral" result. By checking the DNSBL after the known good sources, problems with incorrect listing on the DNSBL are greatly reduced. The RFC5321.MailFrom identity could have additional information in the local-part that cryptographically identifies the mail as coming from an authorized source. In this case, such an SPF record could be used: "v=spf1 mx exists:%{l}._spf_verify.%{d} -all" Then, a specialized DNS server can be set up to serve the _spf_verify subdomain that validates the local-part. Although this requires an extra DNS lookup, this happens only when the email would otherwise be rejected as not coming from a known good source. Note that due to the 63-character limit for domain labels, this approach only works reliably if the local-part signature scheme is guaranteed either to only produce local-parts with a maximum of 63 characters or to gracefully handle truncated local-parts. Similarly, a specialized DNS server could be set up that will rate-limit the email coming from unexpected IP addresses. "v=spf1 mx exists:%{ir}._spf_rate.%{d} -all" SPF allows the creation of per-user policies for special cases. For example, the following SPF record and appropriate wildcard DNS records can be used: "v=spf1 mx redirect=%{l1r+}._at_.%{o}._spf.%{d}" The middle, when email is forwarded (Mediating ADMDs). Forwarding services can solve the problem by rewriting the RFC5321.MailFrom to be in their own domain. This means mail rejected from the external mailbox will have to be forwarded back to the original sender by the forwarding service. Various schemes to do this exist though they vary widely in complexity and resource requirements on the part of the forwarding service. Several popular MTAs can be forced from "alias" semantics to "mailing list" semantics by configuring an additional alias with "owner-" prepended to the original alias name (e.g., an alias of "friends: george@example.com, fred@example.org" would need another alias of the form "owner-friends: localowner"). The end, when email is received (Receiving ADMDs). If the owner of the external mailbox wishes to trust the forwarding service, he can direct the external mailbox's MTA to skip SPF tests when the client host belongs to the forwarding service. Tests against other identities, such as the RFC5321.HELO/.EHLO identity, MAY be used to override a failed test against the RFC5321.MailFrom identity. For larger domains, it might not be possible to have a complete or accurate list of forwarding services used by the owners of the domain's mailboxes. In such cases, whitelists of generally-recognized forwarding services could be employed.

MSPs (Mail Service Providers - Section 2.3) that offer mail services to third-party domains, such as sending of bulk mail, might want to adjust their configurations in light of the authorization check described in this document. If the domain part of the RFC5321.MailFrom identity used for such email uses the domain of one of the MSPs domain, then the provider needs only to ensure that its sending host is authorized by its own SPF record, if any. If the RFC5321.MailFrom identity does not use the MSP's domain, then extra care must be taken. The SPF record format has several options for the third-party domain to authorize the service provider's MTAs to send mail on its behalf. For MSPs, such as ISPs, that have a wide variety of customers using the same MTA, steps should be taken to prevent cross-customer forgery (see ).

Relays are described in Section 2.2.2. The authorization check generally precludes the use of arbitrary MTA relays between sender and receiver of an email message. Within an organization, MTA relays can be effectively deployed. However, for purposes of this document, such relays are effectively transparent. The SPF authorization check is a check between border MTAs of different ADMDs. For mail senders, this means that published SPF records must authorize any MTAs that actually send across the Internet. Usually, these are just the border MTAs as internal MTAs simply forward mail to these MTAs for delivery. The receiving ADMD will generally want to perform the authorization check at the boundary MTAs, including all secondary MXs. Internal MTAs (including MTAs that might serve both as boundary MTAs and internal relays from secondary MXs when they are processing the relayed mail stream) then do not perform the authorization test. To perform the authorization test other than at the boundary, the host that first transferred the message to the receiving ADMD must be determined, which can be difficult to extract from the message header because (a) header fields can be forged or malformed, and (b) there's no standard way to encode that information such that it can be reliably extracted. Testing other than at the boundary is likely to produce unreliable results.

SPF results can be used in combination with other methods to determine the final local disposition (either positive or negative of a message. It can also be considered dispositive on it's own.

SPF pass results can be used in combination with "white lists" of known "good" domains to bypass some or all additional pre-delivery email checks. Exactly which checks and how to determine appropriate white list entries has to be based on local conditions and requirements.

SPF fail results can be used to reject messages during the SMTP transaction based on either RFC5321.MailFrom or RFC5321.HELO/.EHLO identity results. This reduces resource requirements for various content filtering methods and conserves bandwidth since rejection can be done before the SMTP content is transferred. It also gives immediate feedback to the sender who might then be able to resolve the issue. Due to some of the issues described above in this section, SPF based rejection does present some risk of rejecting legitimate email when rejecting based on RFC5321.MailFrom results. SPF fail results can also be used as one input into a larger set of evaluations which might, based on the overall evaluation result in the email being marked negatively in some way (this might be via delivery to a special spam folder, modifying subject lines, or other locally determined means). Developing the details of such an approach have to be based on local conditions and requirements. Using SPF results in this way does not have the advantages of resource conservation and immediate feedback to the sender associated with SMTP rejection, but could produce fewer undesireable rejections in a well designed system. Either general approach can be used as they both leave a clear disposition of emails. They are either delivered in some manner or the sender is notified of the failure. Other dispositions such as "dropping" or deleting email after acceptance are inappropriate because they leave uncertainty and reduce the overall reliabilility and utility of email across the Internet.

The "permerror" result (see ) indicates the SPF processing module at the receiver determined that the retrieved SPF policy record could not be interpreted. This gives no true indication about the authorized use of the data found in the envelope. As with all results, implementers have a choice to make regarding what to do with a message that yields this result. SMTP allows only a few basic options. Rejection of the message is an option, in that it is the one thing a receiver can do to draw attention to the difficulty encountered while protecting itself from messages that do not have a definite SPF result of some kind. However, if the SPF implementation is defective and returns spurious "permerror" results, only the sender is actively notified of the defect (in the form of rejected mail), and not the receiver making use of SPF. The less intrusive handling choice is to deliver the message, perhaps with some kind of annotation of the difficulty encountered and/or logging of a similar nature. However, this will not be desirable to operators that wish to implement SPF checking as strictly as possible, nor is this sort of passive problem reporting typically effective. There is of course the option placing this choice in the hands of the operator rather than the implementer since this kind of choice is often a matter of local policy rather than a condition with a universal solution, but this adds one more piece of complexity to an already non-trivial environment. Both implementers and operators need to be cautious of all choices and outcomes when handling SPF results.

As with most aspects of email, there are a number of ways that malicious parties could use the protocol as an avenue for a Denial-of-Service (DoS) attack. The processing limits outlined in are designed to prevent attacks such as the following: A malicious party could create an SPF record with many references to a victim's domain and send many emails to different SPF clients; those SPF clients would then create a DoS attack. In effect, the SPF clients are being used to amplify the attacker's bandwidth by using fewer bytes in the SMTP session than are used by the DNS queries. Using SPF clients also allows the attacker to hide the true source of the attack. Whereas implementations of check_host() are supposed to limit the number of DNS lookups, malicious domains could publish records that exceed these limits in an attempt to waste computation effort at their targets when they send them mail. Malicious domains could also design SPF records that cause particular implementations to use excessive memory or CPU usage, or to trigger bugs. Malicious parties could send a large volume of mail purporting to come from the intended target to a wide variety of legitimate mail hosts. These legitimate machines would then present a DNS load on the target as they fetched the relevant records. Of these, the case of a third party referenced in the SPF record is the easiest for a DoS attack to effectively exploit. As a result, limits that might seem reasonable for an individual mail server can still allow an unreasonable amount of bandwidth amplification. Therefore, the processing limits need to be quite low. MTAs or other processors SHOULD impose a limit on the maximum amount of elapsed time to evaluate check_host(). Such a limit SHOULD allow at least 20 seconds. If such a limit is exceeded, the result of authorization SHOULD be "temperror".

The RFC5321.MailFrom and RFC5321.HELO/.EHLO identity authorizations must not be construed to provide more assurance than they do. It is entirely possible for a malicious sender to inject a message using his own domain in the identities used by SPF, to have that domain's SPF record authorize the sending host, and yet the message can easily list other identities in its header. Unless the user or the MUA takes care to note that the authorized identity does not match the other more commonly-presented identities (such as the From: header field), the user might be lulled into a false sense of security.

There are two aspects of this protocol that malicious parties could exploit to undermine the validity of the check_host() function: The evaluation of check_host() relies heavily on DNS. A malicious attacker could attack the DNS infrastructure and cause check_host() to see spoofed DNS data, and then return incorrect results. This could include returning "pass" for an <ip> value where the actual domain's record would evaluate to "fail". See for a description of DNS weaknesses. The client IP address, <ip>, is assumed to be correct. A malicious attacker could spoof TCP sequence numbers to make mail appear to come from a permitted host for a domain that the attacker is impersonating.

By definition, SPF policies just map domain names to sets of authorized MTAs, not whole email addresses to sets of authorized users. Although the "l" macro () provides a limited way to define individual sets of authorized MTAs for specific email addresses, it is generally impossible to verify, through SPF, the use of specific email addresses by individual users of the same MTA. It is up to mail services and their MTAs to directly prevent cross-user forgery: based on SMTP AUTH (), users should be restricted to using only those email addresses that are actually under their control (see , Section 6.1). Another means to verify the identity of individual users is message cryptography such as PGP () or S/MIME ().

An SPF compliant receiver gathers information from the SMTP commands it receives and from the published DNS records of the sending domain holder, (e.g., RFC5321.HELO/.EHLO domain name, the RFC5321.MailFrom address from the envelope, and SPF DNS records published by the domain holder). This information, passed to the receiver in the Received-SPF: trace fields, may be returned to the client MTA as an SMTP rejection message. If such an SMTP rejection message is generated, the information from the trace fields must be checked for such problems as invalid characters and excessively long lines. When the authorization check fails, an explanation string could be included in the reject response. Both the sender and the rejecting receiver need to be aware that the explanation was determined by the publisher of the SPF record checked and, in general, not the receiver. The explanation can contain malicious URLs, or it might be offensive or misleading. Explanations returned to sender domains due to "exp" modifiers, (), were generated by the sender policy published by the domain holders themselves. As long as messages are only returned with non-delivery notification () to domains publishing the explanation strings from their own DNS SPF records, the only affected parties are the original publishers of the domain's SPF records. In practice, such non-delivery notifications can be misdirected, such as when an MTA accepts an email and only later generates the notification to a forged address, or when an email forwarder does not direct the bounce back to the original sender.

Checking SPF records causes DNS queries to be sent to the domain owner. These DNS queries, especially if they are caused by the "exists" mechanism, can contain information about who is sending email and likely to which MTA the email is being sent. This can introduce some privacy concerns, which are more or less of an issue depending on local laws and the relationship between the domain owner and the person sending the email.

This document is largely based on the work of Meng Weng Wong, Mark Lentczner, and Wayne Schlitt. Although, as this section acknowledges, many people have contributed to this document, a very large portion of the writing and editing are due to Meng, Mark, and Wayne. This design owes a debt of parentage to by Hadmut Danisch and to by Gordon Fecyk. The idea of using a DNS record to check the legitimacy of an email address traces its ancestry further back through messages on the namedroppers mailing list by Paul Vixie (based on suggestion by Jim Miller) and by David Green . Philip Gladstone contributed the concept of macros to the specification, multiplying the expressiveness of the language and making per-user and per-IP lookups possible. The authors of both this document and would also like to thank the literally hundreds of individuals who have participated in the development of this design. They are far too numerous to name, but they include the following: The participants in the SPFbis working group. The folks on the spf-discuss mailing list. The folks on the SPAM-L mailing list. The folks on the IRTF ASRG mailing list. The folks on the IETF MARID mailing list. The folks on #perl.

Per , the IANA assigned the Resource Record Type and Qtype from the DNS Parameters Registry for the SPF RR type with code 99. The format of this type is identical to the TXT RR . The character content of the record is encoded as . Use of this record type is obsolete for SPF Version 1. IANA is requested to add an annotation to the SPF RRTYPE saying "(OBSOLETE - use TXT)" in the DNS Parameters registry. [NOTE TO RFC EDITOR: (to be changed to " ... has added ..." upon publication)]

Per , the "Received-SPF:" header field is added to the IANA Permanent Message Header Field Registry. The following is the registration template: Header field name: Received-SPF Applicable protocol: mail () Status: Standards Track Author/Change controller: IETF Specification document(s): RFC &rfc.number; [NOTE TO RFC EDITOR: (this document)]

created a new SPF Modifier Registration. IANA is requested to change the reference for the exp and redirect modifiers from to this document. Their status should not be changed.

&rfc1035; &rfc1123; &rfc2119; &rfc3463; &rfc3864; &rfc3986; &rfc4291; &rfc5234; &rfc5321; &rfc5322; &rfc5451; &rfc5598; &rfc5890; American National Standards Institute (formerly United States of America Standards Institute) ANSI X3.4-1968 has been replaced by newer versions with slight modifications, but the 1968 version remains definitive for the Internet. &rfc1034; &rfc1983; &rfc2308; &rfc2782; &rfc3464; &rfc3696; &rfc3833; &rfc3834; &rfc4408; &rfc4632; &rfc4880; &rfc4954; &rfc5617; &rfc5751; &rfc5782; &rfc6409; &rfc6647; &rfc6652; &rfc6686; Work In Progress Work In Progress

This section is normative and any discrepancies with the ABNF fragments in the preceding text are to be resolved in favor of this grammar. See for ABNF notation. Please note that as per this ABNF definition, literal text strings (those in quotes) are case-insensitive. Hence, "mx" matches "mx", "MX", "mX", and "Mx".

; e.g., 2001:DB8::CD30 domain-spec = macro-string domain-end domain-end = ( "." toplabel [ "." ] ) / macro-expand toplabel = ( *alphanum ALPHA *alphanum ) / ( 1*alphanum "-" *( alphanum / "-" ) alphanum ) ; LDH rule plus additional TLD restrictions ; (see [RFC3696], Section 2 for background) alphanum = ALPHA / DIGIT explain-string = *( macro-string / SP ) macro-string = *( macro-expand / macro-literal ) macro-expand = ( "%{" macro-letter transformers *delimiter "}" ) / "%%" / "%_" / "%-" macro-literal = %x21-24 / %x26-7E ; visible characters except "%" macro-letter = "s" / "l" / "o" / "d" / "i" / "p" / "h" / "c" / "r" / "t" / "v" transformers = *DIGIT [ "r" ] delimiter = "." / "-" / "+" / "," / "/" / "_" / "=" name = ALPHA *( ALPHA / DIGIT / "-" / "_" / "." ) header-field = "Received-SPF:" [CFWS] result FWS [comment FWS] [ key-value-list ] CRLF result = "pass" / "fail" / "softfail" / "neutral" / "none" / "temperror" / "permerror" key-value-list = key-value-pair *( ";" [CFWS] key-value-pair ) [";"] key-value-pair = key [CFWS] "=" ( dot-atom / quoted-string ) key = "client-ip" / "envelope-from" / "helo" / "problem" / "receiver" / identity / mechanism / name identity = "mailfrom" ; for the RFC5321.MailFrom identity / "helo" ; for the RFC5321.HELO/.EHLO identity / name ; other identities ALPHA = DIGIT = <0-9 as per [RFC5234]> SP = domain = dot-atom = quoted-string = comment = CFWS = FWS = CRLF = authserv-id = reasonspec = ]]>

These examples are based on the following DNS setup:

; A domain with two mail servers, two hosts ; and two servers at the domain name $ORIGIN example.com. @ MX 10 mail-a MX 20 mail-b A 192.0.2.10 A 192.0.2.11 amy A 192.0.2.65 bob A 192.0.2.66 mail-a A 192.0.2.129 mail-b A 192.0.2.130 www CNAME example.com. ; A related domain $ORIGIN example.org. @ MX 10 mail-c mail-c A 192.0.2.140 ; The reverse IP for those addresses $ORIGIN 2.0.192.in-addr.arpa. 10 PTR example.com. 11 PTR example.com. 65 PTR amy.example.com. 66 PTR bob.example.com. 129 PTR mail-a.example.com. 130 PTR mail-b.example.com. 140 PTR mail-c.example.org. ; A rogue reverse IP domain that claims to be ; something it's not $ORIGIN 0.0.10.in-addr.arpa. 4 PTR bob.example.com.

These examples show various possible published records for example.com and which values if <ip> would cause check_host() to return "pass". Note that <domain> is "example.com". v=spf1 +all any <ip> passes v=spf1 a -all hosts 192.0.2.10 and 192.0.2.11 pass v=spf1 a:example.org -all no sending hosts pass since example.org has no A records v=spf1 mx -all sending hosts 192.0.2.129 and 192.0.2.130 pass v=spf1 mx:example.org -all sending host 192.0.2.140 passes v=spf1 mx mx:example.org -all sending hosts 192.0.2.129, 192.0.2.130, and 192.0.2.140 pass v=spf1 mx/30 mx:example.org/30 -all any sending host in 192.0.2.128/30 or 192.0.2.140/30 passes v=spf1 ptr -all sending host 192.0.2.65 passes (reverse DNS is valid and is in example.com) sending host 192.0.2.140 fails (reverse DNS is valid, but not in example.com) sending host 10.0.0.4 fails (reverse IP is not valid) v=spf1 ip4:192.0.2.128/28 -all sending host 192.0.2.65 fails sending host 192.0.2.129 passes

These examples show the effect of related records: example.org: "v=spf1 include:example.com include:example.net -all" This record would be used if mail from example.org actually came through servers at example.com and example.net. Example.org's designated servers are the union of example.com's and example.net's designated servers. la.example.org: "v=spf1 redirect=example.org" ny.example.org: "v=spf1 redirect=example.org" sf.example.org: "v=spf1 redirect=example.org" These records allow a set of domains that all use the same mail system to make use of that mail system's record. In this way, only the mail system's record needs to be updated when the mail setup changes. These domains' records never have to change.

Imagine that, in addition to the domain records listed above, there are these:

$ORIGIN _spf.example.com. mary.mobile-users A 127.0.0.2 fred.mobile-users A 127.0.0.2 15.15.168.192.joel.remote-users A 127.0.0.2 16.15.168.192.joel.remote-users A 127.0.0.2

The following records describe users at example.com who mail from arbitrary servers, or who mail from personal servers. example.com:

v=spf1 mx include:mobile-users._spf.%{d} include:remote-users._spf.%{d} -all

mobile-users._spf.example.com:

v=spf1 exists:%{l1r+}.%{d}

remote-users._spf.example.com:

v=spf1 exists:%{ir}.%{l1r+}.%{d}

Say that your sender policy requires both that the IP address is within a certain range and that the reverse DNS for the IP matches. This can be done several ways, including the following:

example.com. SPF ( "v=spf1 " "-include:ip4._spf.%{d} " "-include:ptr._spf.%{d} " "+all" ) ip4._spf.example.com. SPF "v=spf1 -ip4:192.0.2.0/24 +all" ptr._spf.example.com. SPF "v=spf1 -ptr +all"

This example shows how the "-include" mechanism can be useful, how an SPF record that ends in "+all" can be very restrictive, and the use of De Morgan's Law.

Changes since RFC 4408 (to be removed prior to publication) Moved to standards track Authors updated IESG Note regarding experimental use replaced with discussion of results Process errata: Add %v macro to ABNF grammar Replace "uric" by "unreserved" Recommend an SMTP reply code for optional permerror rejections Correct syntax in Received-SPF examples Fix unknown-modifier clause is too greedy in ABNF Correct use of empty domain-spec on exp modifier Fix minor typo errata Convert to spfbis working group draft, draft-ietf-spfbis-4408bis-00 Addressed Ticket #1, RFC 4408 Section 2.5.6 - Temporary errors by giving the option to turn repeated SERVFAIL into permerror and adding RFC 2308 reference. Clarified text about IPv4 mapped addresses to resolve test suite ambiguity Clarified ambiguity about result when more than 10 "mx" or "ptr" records are returned for lookup to specify permerror. This resolves one of the test suite ambiguities Made all references to result codes lower case per issue #7 Adjusted section 2.2 Requirement to check mail from per issue #15 Added missing "v" element in macro-letter in the collected ABNF per issue #16 - section 8.1 was already fixed in the pre-WG draft Marked ptr and "p" macro deprecated/SHOULD NOT use per issue #27 Expunged lower case may from the draft per issue #8 Expunged "x-" name as an obsolete concept Updated obslete references: RFC2821 to RFC5321, RFC2822 to RFC5322, and RFC4234 to RFC5234 Refer to RFC6647 to describe greylisting instead of trying to describe it directly. Updated informative references to the current versions. Added definition for deprecated since there are questions. Start to rework section 9 with some RFC5598 terms. Added mention of RFC 6552 feedback reports in section 9. Added draft-ietf-spfbis-experiment as an informational reference. Drop Type SPF. Try and clarify informational nature of RFC3696 Fix ABNF nits and add missing definitions per Bill's ABNF checker. Make DNS lookup time limit SHOULD instead of MAY. Reorganize and clarify processing limits. Move hard limits to new section 4.6.4, Evaluation Limits. Move advice to non-normative section 9. Removed paragraph in section 10.1 about limiting total data volumes as it is unused (and removable per the charter) and serves no purpose (it isn't something that actually can be implemented in any reasonable way). Added text and figures from Alessandro Vesely in section 9.1 to better explain DNS resource limits. Multiple editorial fixes from Murray Kucherawy's review. Also based on Murray's review, reworked SMTP identity definitions and made RFC 5598 a normative reference instead of informative. This is a downref that will have to be mentioned in the last call. Added RFC 3834 as an informative reference about backscatter. Added IDN requirements and normative reference to RFC 5890 to deal with the question "like DKIM did it.: Added informative reference to RFC 4632 for CIDR and use CIDR prefix length instead of CIDR-length to match its terminology. Added RFC 5782 informative reference on DNSxLs to support improving the exists description. Added text on creating a Authentication-Results header field that matches the Received-SPF header field information and added a normative reference to RFC 5451. Added informative reference to RFC 2782 due to SRV mention. Added informative reference to RFC 3464 due to DSN mention. Added informative reference to RFC 5617 for it's DNS wildcard use. Added informative reference to RFC 5782 to enhance the explanation of how the exists mechanism works. Clarified the intended match/no-match method. Added new sections on Receiver policy for SPF pass, fail, and permerror. Added new section 9 discussion on treatment of bounces and the significance of HELO records. Added request to IANA to update the SPF modifier registry.