Mail::SPF::Query.3pm

Langue: en

Version: 2006-02-26 (mandriva - 01/05/08)

Section: 3 (Bibliothèques de fonctions)

NAME

Mail::SPF::Query - query Sender Policy Framework for an IP,email,helo

VERSION

1.999.1

SYNOPSIS

     my $query = new Mail::SPF::Query (ip => "127.0.0.1", sender=>'foo@example.com', helo=>"somehost.example.com", trusted=>0, guess=>0);
     my ($result,           # pass | fail | softfail | neutral | none | error | unknown [mechanism]
         $smtp_comment,     # "please see http://www.openspf.org/why.html?..."  when rejecting, return this string to the SMTP client
         $header_comment,   # prepend_header("Received-SPF" => "$result ($header_comment)")
         $spf_record,       # "v=spf1 ..." original SPF record for the domain
        ) = $query->result();
 
     if    ($result eq "pass") { "Domain is not forged. Apply RHSBL and content filters." }
     elsif ($result eq "fail") { "Domain is forged. Reject or save to spambox." }
 
 

ABSTRACT

The SPF protocol relies on sender domains to describe their designated outbound mailers in DNS. Given an email address, Mail::SPF::Query determines the legitimacy of an SMTP client IP address.

DESCRIPTION

There are two ways to use Mail::SPF::Query. Your choice depends on whether the domains your server is an MX for have secondary MXes which your server doesn't know about.

The first and more common style, calling ->result(), is suitable when all mail is received directly from the originator's MTA. If the domains you receive do not have secondary MX entries, this is appropriate. This style of use is outlined in the SYNOPSIS above. This is the common case.

The second style is more complex, but works when your server receives mail from secondary MXes. This performs checks as each recipient is handled. If the message is coming from a valid MX secondary for a recipient, then the SPF check is not performed, and a ``pass'' response is returned right away. To do this, call "result2()" and "message_result2()" instead of "result()".

If you do not know what a secondary MX is, you probably don't have one. Use the first style.

You can try out Mail::SPF::Query on the command line with the following command:

     perl -MMail::SPF::Query -le 'print for Mail::SPF::Query->new(
         helo => shift, ipv4 => shift, sender => shift)->result' \
         helohost.example.com 1.2.3.4 user@example.com
 
 

BUGS

Mail::SPF::Query tries to implement the SPF specification (see ``SEE ALSO'') as close as reasonably possible given that M:S:Q has been the very first SPF implementation and has changed with the SPF specification over time. As a result, M:S:Q has various known deficiencies that cannot be corrected with reasonably little effort:
Unable to query HELO and MAIL FROM separately. M:S:Q is not designed to support the separate querying of the HELO and MAIL FROM identities. Passing the HELO identity as the "sender" argument for a stand-alone HELO check might generally work but could yield unexpected results.
No IPv6 support. IPv6 is not supported. "ip6" mechanisms in SPF records and everywhere else are simply ignored.
Result explanation may be inappropriate for local policy results. If a query result was caused by anything other than a real SPF record (i.e. local policy, overrides, fallbacks, etc.), and no custom "default_explanation" was specified, the domain's explanation or M:S:Q's hard-coded default explanation will still be returned. Be aware that in this case the explanation may not correctly explain the reason for such an artificial result.

NON-STANDARD FEATURES

Also due to its long history, M:S:Q does have some legacy features that are not parts of the official SPF specification, most notably best guess processing and trusted forwarder accreditation checking. Please be careful when using these non-standard features or when reproducing them in your own SPF implementation, as they may cause unexpected results.

METHODS


"Mail::SPF::Query->new()"


Mail::SPF::Query->new()

     my $query = eval { new Mail::SPF::Query (
         ip          => '127.0.0.1',
         sender      => 'foo@example.com',
         helo        => 'host.example.com',
 
         # Optional parameters:
         debug       => 1, debuglog => sub { print STDERR "@_\n" },
         local       => 'extra mechanisms',
         trusted     => 1,                   # do trusted forwarder processing
         guess       => 1,                   # do best guess if no SPF record
         default_explanation => 'Please see http://spf.my.isp/spferror.html for details',
         max_lookup_count    => 10,          # total number of SPF includes/redirects
         sanitize    => 0,                   # do not sanitize all returned strings
         myhostname  => 'foo.example.com',   # prepended to header_comment
         override    => {   'example.net' => 'v=spf1 a mx -all',
                          '*.example.net' => 'v=spf1 a mx -all' },
         fallback    => {   'example.org' => 'v=spf1 a mx -all',
                          '*.example.org' => 'v=spf1 a mx -all' }
     ) };
 
     if ($@) { warn "bad input to Mail::SPF::Query: $@" }
 
 

Set "trusted=>1" to turned on "trusted-forwarder.org" accreditation checking. The mechanism "include:spf.trusted-forwarder.org" is used just before a "-all" or "?all". The precise circumstances are somewhat more complicated, but it does get the case of "v=spf1 -all" right --- i.e. "trusted-forwarder.org" is not checked. This is a non-standard feature.

Set "guess=>1" to turned on automatic best guess processing. This will use the best_guess SPF record when one cannot be found in the DNS. Note that this can only return "pass" or "neutral". The "trusted" and "local" flags also operate when the best_guess is being used. This is a non-standard feature.

Set "local=>'include:local.domain'" to include some extra processing just before a "-all" or "?all". The local processing happens just before the trusted forwarder processing. This is a non-standard feature.

Set "default_explanation" to a string to be used if the SPF record does not provide a specific explanation. The default value will direct the user to a page at www.openspf.org with the following message:

     Please see http://www.openspf.org/why.html?sender=%{S}&ip=%{I}&receiver=%{R}
 
 

Note that the string has macro substitution performed.

Set "sanitize" to 0 to get all the returned strings unsanitized. Alternatively, pass a function reference and this function will be used to sanitize the returned values. The function must take a single string argument and return a single string which contains the sanitized result.

Set "debug=>1" to watch the queries happen.

Set "override" to define SPF records for domains that do publish but which you want to override anyway. Wildcards are supported. This is a non-standard feature.

Set "fallback" to define ``pretend'' SPF records for domains that don't publish them yet. Wildcards are supported. This is a non-standard feature.

Note: domain name arguments to override and fallback need to be in all lowercase.

"$query->result()"


$query->result()

     my ($result, $smtp_comment, $header_comment, $spf_record, $detail) = $query->result();
 
 

$result will be one of "pass", "fail", "softfail", "neutral", "none", "error" or "unknown [...]":

"pass"
The client IP address is an authorized mailer for the sender. The mail should be accepted subject to local policy regarding the sender.
"fail"
The client IP address is not an authorized mailer, and the sender wants you to reject the transaction for fear of forgery.
"softfail"
The client IP address is not an authorized mailer, but the sender prefers that you accept the transaction because it isn't absolutely sure all its users are mailing through approved servers. The "softfail" status is often used during initial deployment of SPF records by a domain.
"neutral"
The sender makes no assertion about the status of the client IP.
"none"
There is no SPF record for this domain.
"error"
The DNS lookup encountered a temporary error during processing.
"unknown [...]"
The domain has a configuration error in the published data or defines a mechanism that this library does not understand. If the data contained an unrecognized mechanism, it will be presented following ``unknown''. You should test for unknown using a regexp "/^unknown/" rather than "eq "unknown"".

Results are cached internally for a default of 120 seconds. You can call "->result()" repeatedly; subsequent lookups won't hit your DNS.

"smtp_comment" should be displayed to the SMTP client.

"header_comment" goes into a "Received-SPF" header, like so:

     Received-SPF: $result ($header_comment)
 
 

"spf_record" shows the original SPF record fetched for the query. If there is no SPF record, it is blank. Otherwise, it will start with "v=spf1" and contain the SPF mechanisms and such that describe the domain.

Note that the strings returned by this method (and most of the other methods) are (at least partially) under the control of the sender's domain. This means that, if the sender is an attacker, the contents can be assumed to be hostile. The various methods that return these strings make sure that (by default) the strings returned contain only characters in the range 32 - 126. This behavior can be changed by setting "sanitize" to 0 to turn off sanitization entirely. You can also set "sanitize" to a function reference to perform custom sanitization. In particular, assume that "smtp_comment" might contain a newline character.

"detail" is a hash of all the foregoing result elements, plus extra data returned by the SPF result.

Why the weird duplication? In the beginning, "result()" returned only one value, the $result. Then $smtp_comment and $header_comment came along. Then $spf_record. Past a certain number of positional results, it makes more sense to have a hash. But we didn't want to break backwards compatibility, so we just declared that the fifth result would be a hash and future return value would go in there.

The keys of the hash are:

     result
     smtp_comment
     header_comment
     header_pairs
     spf_record
     modifiers
 
 

"$query->result2()"


$query->result2()

     my ($result, $smtp_comment, $header_comment, $spf_record) = $query->result2('recipient@domain', 'recipient2@domain');
 
 

"result2()" does everything that "result()" does, but it first checks to see if the sending system is a recognized MX secondary for the recipient(s). If so, then it returns "pass" and does not perform the SPF query. Note that the sending system may be a MX secondary for some (but not all) of the recipients for a multi-recipient message, which is why result2 takes an argument list. See also "message_result2()".

This is a non-standard feature. This feature is also deprecated, because exemption of trusted relays, such as secondary MXes, should really be performed by the software that uses this library before doing an SPF check.

$result will be one of "pass", "fail", "neutral [...]", or "unknown". See "result()" above for meanings.

If you have secondary MXes and if you are unable to explicitly white-list them before SPF tests occur, you can use this method in place of "result()", calling it as many times as there are recipients, or just providing all the recipients at one time.

"smtp_comment" can be displayed to the SMTP client.

For example:

     my $query = new Mail::SPF::Query (ip => "127.0.0.1",
                                       sender=>'foo@example.com',
                                       helo=>"somehost.example.com");
 
     ...
 
     my ($result, $smtp_comment, $header_comment);
 
     ($result, $smtp_comment, $header_comment) = $query->result2('recip1@example.com');
     # return suitable error code based on $result eq 'fail' or not
 
     ($result, $smtp_comment, $header_comment) = $query->result2('recip2@example.org');
     # return suitable error code based on $result eq 'fail' or not
 
     ($result, $smtp_comment, $header_comment) = $query->message_result2();
     # return suitable error if $result eq 'fail'
     # prefix message with "Received-SPF: $result ($header_comment)"
 
 

"$query->message_result2()"


$query->message_result2()

     my ($result, $smtp_comment, $header_comment, $spf_record) = $query->message_result2();
 
 

"message_result2()" returns an overall status for the message after zero or more calls to "result2()". It will always be the last status returned by "result2()", or the status returned by "result()" if "result2()" was never called.

$result will be one of "pass", "fail", "neutral [...]", or "error". See "result()" above for meanings.

"$query->best_guess()"


$query->best_guess()

     my ($result, $smtp_comment, $header_comment) = $query->best_guess();
 
 

When a domain does not publish an SPF record, this library can produce an educated guess anyway.

It pretends the domain defined A, MX, and PTR mechanisms, plus a few others. The default set of directives is

     a/24 mx/24 ptr
 
 

That default set will return either ``pass'' or ``neutral''.

If you want to experiment with a different default, you can pass it as an argument: "$query->best_guess("a mx ptr")"

This is a non-standard feature. This method is also deprecated. You should set "guess=>1" on the "new()" method instead.

"$query->trusted_forwarder()"


$query->trusted_forwarder()

     my ($result, $smtp_comment, $header_comment) = $query->best_guess();
 
 

It is possible that the message is coming through a known-good relay like "acm.org" or "pobox.com". During the transitional period, many legitimate services may appear to forge a sender address: for example, a news website may have a ``send me this article in email'' link.

The "trusted-forwarder.org" domain is a white-list of known-good hosts that either forward mail or perform benign envelope sender forgery:

     include:spf.trusted-forwarder.org
 
 

This will return either ``pass'' or ``neutral''.

This is a non-standard feature. This method is also deprecated. You should set "trusted=>1" on the "new()" method instead.

"$query->sanitize('string')"


$query->sanitize('string')

This applies the sanitization rules for the particular query object. These rules are controlled by the "sanitize" parameter to the c<new()> method.

"strict_sanitize('string')"


strict_sanitize('string')

This ensures that all the characters in the returned string are printable. All whitespace is converted into spaces, and all other non-printable characters are converted into question marks. This is probably over-aggressive for many applications.

This function is used by default when the "sanitize" option is passed to the "new()" method.

This function is not a class method.

"$query->debuglog()"


$query->debuglog()

Subclasses may override this with their own debug logger. "Log::Dispatch" is recommended.

Alternatively, pass the "new()" constructor a "debuglog => sub { ... }" callback, and we'll pass debugging lines to that.

WARNINGS

Mail::Query::SPF should only be used at the point where messages are received from the Internet. The underlying assumption is that the sender of the e-mail is sending the message directly to you or one of your secondary MXes. If your MTA does not have an exhaustive list of secondary MXes, then the "result2()" and "message_result2()" methods can be used. These methods take care to permit mail from secondary MXes.

AUTHORS

Meng Weng Wong <mengwong+spf@pobox.com>, Philip Gladstone, Julian Mehnle <julian@mehnle.net>

SEE ALSO

About SPF: <http://www.openspf.org>

Mail::SPF::Query: <http://search.cpan.org/dist/Mail-SPF-Query>

The latest release of the SPF specification: <http://www.openspf.org/spf-classic-current.txt>