Thomas Hurst <tom.hurst / clara.net> writes:

> After seeing the huge ugly mess of procmail's code, I'm feeling a
> compulsion to change.

http://www.lickey.com/rubymail/ is the result of similar feelings on
my part.  It is a work in progress.

Another Ruby solution is Scarlet.  It is much more mature but is
mainly documented in Japanese so I don't know too many details.  :-)
It does lack one thing I like about my solution -- in RubyMail the
mail filter script is written in pure Ruby.


> The most "standard" alternative seems to be maildrop, which at least
> looks fairly clean and has a much more readable filter language.

True, but it is a crippled little filtering language akin to procmail.
I.e. to do anything complex you have to shell out to an external
command.


> But it still lacks some features I want, such as the ability to
> queue messages and perform higher level comparisons on them to, for
> instance, intelligently deal with dupes in group replies (try
> filtering on X-Mailing-List: headers and tell me to use formail and
> a message id cache..)

Yeah, this is often the kind of stuff you often want or need a "real"
language to do.


> Then I started thinking about how to use ruby as a filter language;

[...]

> queue = Queue.new(20, '3 min')
> queue.prune_dupes
> queue.run do |mail|

Not sure what you mean here...does this read form your inbox or
something?

>         mail.cc('Backup')
>         if list = MailingList.filter(mail)
>                 mail.deliver('lists/' + list)
>         end
>         mail.filter('/usr/local/bin/spamassassin -P')
>         if mail.header.match(/^X-Spam-Flag: Yes/)
>                 mail.deliver('SPAM')
>         end
>         mail.deliver('Inbox')

Yeah, I have basic mail handling classes (Mail::Message,
Mail::Header), basic classes use to parse messages (Mail::Parser) and,
soon, a basic class used to print a message to an output stream
(Mail::Printer -- class name is undecided!).  In the design of these
classes I'm finding I'm stealing lots of ideas from Python's email
library and to a lesser extent Perl's Mail::Internet and MIME-tools
packages, and Ruby's TMail.

Then I have a Mail::LDA (local delivery agent) class that makes it
easy to write mail filters without worrying about a random exception
in your code causing a bounce.

Then there is an rdeliver.rb script that uses Mail::LDA and friends to
parse a message from stdin.  It then creates a Deliver class, and
evals the user's ~/.rdeliver file within the context of that class,
which is responsible for defining Deliver#main.  The script then
instantiates a Deliver object can calls Deliver#main.

The simplest possible .rdeliver file is:

    def main
        lda.save("inbox")
    end

This defines a Deliver#main method that saves the message to a unix
mbox file called "inbox" in the user's home directory.

Here is some snippets from my .rdeliver file.  There is some stuff
here that could be factored out into nicer pieces/modules.  But you'll
get the idea.  Remember, this is evaluated in the context of the
Deliver class, so each def here adds a method to the Deliver class.

def spool_save(folder, continue = false)
  lda.save(".incoming/#{folder}.spool", continue)
end

def save_if_list(list, folder)
  name, domain = list.split(/@/).collect { |s| Regexp.quote(s) }
  if h.match?(/^((Resent-)?Sender|Errors-To|X-Loop|(X-)?Mailing-List)$/i, /(owner-)?#{name}(-request|-help|-admin|-bounce|-errors|-owner)?@#{domain}/im) ||
      h.match?(/(List-Id|X-Mailing-List)$/i, /<#{name}.#{domain}>/im) ||
      h.match?("x-ml-name", /^\s*#{name}([^ \t]|$)/im) ||
      h.match?(/^(to|cc)$/i, /#{name}@#{domain}/im)
    spool_save("list.#{folder}")
  end
end

# This function checks if a given +ip+ is listed by the indicated DNS
# blackhole list +service+.  It returns a text string indicating why
# the ip was blocked if listed, or nil if not listed.
#
# This function relies on the 'host' external command.
def dnsbl(ip, service)
  if ip =~ /\A\d+\.\d+\.\d+\.\d+\z/
    ip.untaint
  end
  raise "bad ip #{ip.inspect}" if ip.tainted?
  query = ip.split(/\./).reverse.join('.') + '.' + service
  result = `host -t txt #{query} 2> /dev/null`
  return nil unless $? == 0
  unless result =~ /\"(.*)\"/m
    raise "can't find text record in ${result.inspect}"
  end
  return "#{service} lists ip #{ip}: " + $1
end

def main

[...]

  save_if_list('briefs / lists.freeswan.org', 'freeswan')
  save_if_list('announce / lists.freeswan.org', 'freeswan')
[...]
  save_if_list('zsh-workers / sunsite.dk', 'zsh')
  save_if_list('zsh-users / sunsite.dk', 'zsh')

  # backup everything
  lda.save('.incoming/backup-' + Time.now.strftime('%Y-%V'), true)

  # Check if we got this through bigfoot and bounce it if it looks
  # suspicious.
  if h.match?('delivered-to', /matt\+bigfoot@lickey\.com/im) &&
      h.match?('received', /from.*by.*bigfoot\.com.*LiteMail/im)
    temp = h.match('received', /from.*by.*bigfoot\.com.*LiteMail/im)[0]
    if temp =~ /\(\[([\d\.]+)\]\)/
      ip = $1
      raise if defined? blocked
      lda.log(1, "bigfoot ip is #{ip}")
      blocked ||= dnsbl(ip, 'relays.ordb.org')
      blocked ||= dnsbl(ip, 'inputs.orbz.org')
      blocked ||= dnsbl(ip, 'bl.spamcop.net')
      lda.reject(blocked) unless blocked.nil?
    end
  end

[...]

  if h.match?('from', /matt@lickey\.com/im) &&
      h.match?('subject', /testme bounce/im)
    lda.reject("You want to test a bounce?  I'll give you a bounce!")
  end

[...]

  if h.match?('from', /Cron Daemon|root@lickey\.com|uucp@lickey\.com|cfengine@lickey\.com/im)
    spool_save("daemons")
  end

  # FIXME: need a forward mechanism
  if h.match?('from', /relaytest / lickey.com/im) &&
      h.match?('subject', /open relay test for/im)
    lda.pipe("/usr/sbin/sendmail -oi qspam / orbz.org relays / ordb.org")
  end

[...]

  spool_save("inbox")
end


-- 
matt