blog of zachary "za3k" vance

I’m posting my email setup here. The end result will:

Use Postfix for SMTP
Use Dovecot for IMAP and authentication
Store usernames, email forwards, and passwords in a Postgres SQL database
Only be accessible over encrypted channels
Pass all common spam checks
Support SMTP sending and IMAP email checking. I did not include POP3 because I don’t use it, but it should be easy to add
NOT add spam filtering or web mail (this article is long enough as it is, maybe in a follow-up)

Note: My set up is pretty standard, except that rDNS for smtp.za3k.com resolves to za3k.com because I only have one IP. You may need to change your hostnames if you’re using mail.example.com or smtp.example.com.

On to the install!

Install debian packages

sudo apt-get install postfix # Postfix \
      dovecot-core dovecot-imapd dovecot-lmtpd # Dovecot \
      postgresql dovecot-pgsql postfix-pgsql # Postgres \
      opendkim opendkim-tools # DKIM

Set up security. smtp.za3k.com cert is at /etc/certs/zak3.com.pem, the key is at /etc/ssl/private/smtp.za3k.com.key. dhparams for postfix are at /etc/postfix/dhparams.pem. (If you need a certificate and don’t know how to get one, you can read Setting up SSL certificates using StartSSL)

Install Postfix

# /etc/postfix/master.cf
smtp       inet  n       -       -       -       -       smtpd
submission inet  n       -       -       -       -       smtpd
  -o syslog_name=postfix/submission
  -o smtpd_tls_security_level=encrypt
  -o smtpd_sasl_auth_enable=yes
  -o smtpd_reject_unlisted_recipient=no
  -o milter_macro_daemon_name=ORIGINATING

# /etc/postfix/main.cf additions
# TLS parameters
smtpd_tls_cert_file=/etc/ssl/certs/smtp.za3k.com.pem
smtpd_tls_key_file=/etc/ssl/private/smtp.za3k.com.key
smtpd_use_tls=yes
smtpd_tls_mandatory_protocols=!SSLv2,!SSLv3
smtp_tls_mandatory_protocols=!SSLv2,!SSLv3
smtpd_tls_protocols=!SSLv2,!SSLv3
smtp_tls_protocols=!SSLv2,!SSLv3
smtpd_tls_exclude_ciphers = aNULL, eNULL, EXPORT, DES, RC4, MD5, PSK, aECDH, EDH-DSS-DES-CBC3-SHA, EDH-RSA-DES-CDC3-SHA, KRB5-DE5, CBC3-SHA

# Relay and recipient settings
myhostname = za3k.com
myorigin = /etc/mailname
mydestination = za3k.com, smtp.za3k.com, localhost.com, localhost
relayhost =
mynetworks_style = host
mailbox_size_limit = 0
inet_interfaces = all
smtpd_relay_restrictions = permit_mynetworks,
  permit_sasl_authenticated,
  reject_unauth_destination

alias_maps = hash:/etc/aliases
local_recipient_maps = $alias_maps
mailbox_transport = lmtp:unix:private/dovecot-lmtp

Install Dovecot

# /etc/dovecot/dovecot.cf
mail_privileged_group = mail # Local mail
disable_plaintext_auth = no

protocols = imap

ssl=required
ssl_cert = </etc/ssl/certs/imap.za3k.com.pem
ssl_key = </etc/ssl/private/imap.za3k.com.key

# IMAP Folders
namespace {
 inbox = yes
 mailbox Trash {
 auto = create
 special_use = \Trash
 }
 mailbox Drafts {
 auto = no
 special_use = \Drafts
 }
 mailbox Sent {
 auto = subscribe
 special_use = \Sent
 }
 mailbox Spam {
 auto = subscribe
 special_use = \Junk
 }
}

# Expunging / deleting mail should FAIL, use the lazy_expunge plugin for this
namespace {
 prefix = .EXPUNGED/
 hidden = yes
 list = no
 location = maildir:~/expunged
}
mail_plugins = $mail_plugins lazy_expunge
plugin {
 lazy_expunge = .EXPUNGED/
}

# /etc/postfix/main.cf
# SASL authentication is done through Dovecot to let users relay mail
smtpd_sasl_type = dovecot
smtpd_sasl_path = private/auth

Set up the database and virtual users. Commands

# Create the user vmail for storing virtual mail
# vmail:x:5000:5000::/var/mail/vmail:/usr/bin/nologin
groupadd -g 5000 vmail
mkdir /var/mail/vmail
useradd -M -d /var/mail/vmail --shell=/usr/bin/nologin -u 5000 -g vmail vmail
chown vmail:vmail /var/mail/vmail
chmod 700 /var/mail/vmail

psql -U postgres
; Set up the users
CREATE USER 'postfix' PASSWORD 'XXX';
CREATE USER 'dovecot' PASSWORD 'XXX';

; Create the database
CREATE DATABASE email;
\connect email

; Set up the schema 

CREATE TABLE aliases (
    alias text NOT NULL,
    email text NOT NULL
);

CREATE TABLE users (
    username text NOT NULL,
    domain text NOT NULL,
    created timestamp with time zone DEFAULT now(),
    password text NOT NULL
);

REVOKE ALL ON TABLE aliases FROM PUBLIC;
GRANT ALL ON TABLE aliases TO postfix;
GRANT ALL ON TABLE aliases TO dovecot;

REVOKE ALL ON TABLE users FROM PUBLIC;
GRANT ALL ON TABLE users TO dovecot;
GRANT ALL ON TABLE users TO postfix;

# /etc/dovecot/dovecot.conf
# Since we're giving each virtual user their own directory under /var/mail/vmail, just use that directly and not a subdirectory
mail_location = maildir:~/

# /etc/dovecot/dovecot-sql.conf defines the DB queries used for authorization
passdb {
  driver = sql
  args = /etc/dovecot/dovecot-sql.conf
}
userdb {
  driver = prefetch
}
userdb {
  driver = sql
  args = /etc/dovecot/dovecot-sql.conf
}

# /etc/postfix/main.cf
pgsql:/etc/postfix/pgsql-virtual-aliases.cf
local_recipient_maps = pgsql:/etc/postfix/pgsql-virtual-mailbox.cf

# /etc/postfix/pgsql-virtual-aliases.cf
# hosts = localhost
user = postfix
password = XXXXXX
dbname = email

query = SELECT email FROM aliases WHERE alias='%s'

# /etc/postfix/pgsql-virtual-mailbox.cf
# hosts = localhost
user = postfix
password = XXXXXX
dbname = email

query = SELECT concat(username,'@',domain,'/') as email FROM users WHERE username='%s'

# /etc/dovecot/dovecot-sql.conf
driver = pgsql
connect = host=localhost dbname=email user=dovecot password=XXXXXX
default_pass_scheme = SHA512
password_query = SELECT \
  CONCAT(username,'@',domain) as user, \
  password, \
  'vmail' AS userdb_uid, \
  'vmail' AS userdb_gid, \
  '/var/mail/vmail/%u' as userdb_home \
  FROM users \
  WHERE concat(username,'@',domain) = '%u';

user_query = SELECT username, \
  CONCAT('maildir:/var/mail/vmail/',username,'@',domain) as mail, \
  '/var/mail/vmail/%u' as home, \
  'vmail' as uid, \
  'vmail' as gid \
  FROM users \
  WHERE concat(username,'@',domain) = '%u';

Set up users. Example user creation:

# Generate a password
$ doveadm pw -s sha512 -r 100
Enter new password: ...
Retype new password: ...
{SHA512}.............................................................==

psql -U dovecot -d email
; Create a user za3k@za3k.com
mail=# INSERT INTO users (
    username,
    domain,
    password
) VALUES (
    'za3k',
    'za3k.com'
    '{SHA512}.............................................................==',
);

Set up aliases/redirects. Example redirect creation:

psql -U dovecot -d email
; Redirect mail from foo@example.com to bar@example.net
mail=# INSERT INTO users ( email, alias ) VALUES (
    'bar@example.net',
    'foo@example.com'
);

Test setup locally by hand. Try using TELNET. Test remote setup using STARTSSL. This is similar to the previous step, but to start the connection use:
```
openssl s_client -connect smtp.za3k.com:587 -starttls smtp
```
Make sure to test email to addresses at your domain or that you’ve set up (final destination), and emails you’re trying to send somewhere else (relay email)

A small digression: port 25 is used for unencrypted email and support STARTTLS, 587 is used for STARTTLS only, and 465 (obsolete) is used for TLS. My ISP, Comcast, blocks access to port 25 on outgoing traffic.
Make sure you’re not running an open relay at http://mxtoolbox.com/diagnostic.aspx
Set your DNS so that the MX record points at your new mailserver. You’ll probably want a store and forward backup mail server (mine is provided by my registrar). Email should arrive at your mail server from now on. This is the absolute minimum setup. Everything from here on is to help the community combat spam (and you not to get blacklisted).

Set up DKIM (DomainKeys Identified Mail). DKIM signs outgoing mail to show that it’s from your server, which helps you not get flagged as spam.
None of these files or folders exist to begin with in debian.

# Add to /etc/opendkim.conf
KeyTable                /etc/opendkim/KeyTable
SigningTable            /etc/opendkim/SigningTable
ExternalIgnoreList      /etc/opendkim/TrustedHosts
InternalHosts           /etc/opendkim/TrustedHosts
LogWhy yes

# /etc/opendkim/TrustedHosts
127.0.0.1
[::1]
localhost
za3k.com
smtp.za3k.com

mkdir -p /etc/opendkim/keys/za3k.com
cd /etc/opendkim/keys/za3k.com
opendkim-genkey -s default -d za3k.com
chown opendkim:opendkim default.private

# /etc/opendkim/KeyTable
default._domainkey.za3k.com za3k.com:default:/etc/opendkim/keys/za3k.com/default.private

# /etc/opendkim/SigningTable
za3k.com default._domainkey.za3k.com

Display the DNS public key to set in a TXT record with:

# sudo cat /etc/opendkim/keys/za3k.com/default.txt
default._domainkey      IN      TXT     ( "v=DKIM1; k=rsa; "
          "p=MIGfMA0GCSqGSIb3DQEBAQUAA4GNADCBiQKBgQCggdv3OtQMek/fnu+hRrHYZTUcpUFcSGL/+Sbq+GffR98RCgabx/jjPJo3HmqsB8czaXf7yjO2UiSN/a8Ae6/yu23d7hyTPUDacatEM+2Xc4/zG+eAlAMQOLRJeo3z53sNiq0SmJET6R6yH4HCv9VkuS0TQczkvME5hApft+ZedwIDAQAB" )  ; ----- DKIM

# My registrar doesn't support this syntax so it ends up looking like: 
$ dig txt default._domainkey.za3k.com txt
default._domainkey.za3k.com. 10800 IN   TXT     "v=DKIM1\; k=rsa\; p=MIGfMA0GCSqGSIb3DQEBAQUAA4GNADCBiQKBgQCggdv3OtQMek/fnu+hRrHYZTUcpUFcSGL/+Sbq+GffR98RCgabx/jjPJo3HmqsB8czaXf7yjO2UiSN/a8Ae6/yu23d7hyTPUDacatEM+2Xc4/zG+eAlAMQOLRJeo3z53sNiq0SmJET6R6yH4HCv9VkuS0TQczkvME5hApft+ZedwIDAQAB"

# Uncomment in /etc/default/opendkim
SOCKET="inet:12345@localhost" # listen on loopback on port 12345

# /etc/postfix/main.cf
# DKIM
milter_default_action = accept
milter_protocol = 6
smtpd_milters = inet:localhost:12345
non_smtpd_milters = inet:localhost:12345

Set up SPF (Sender Policy Framework). SPF explains to other services which IPs can send email on your behalf. You can set up whatever policy you like. A guide to the syntax is at: http://www.openspf.org/SPF_Record_Syntax. Mine is
```
@ 10800 IN TXT "v=spf1 +a:za3k.com +mx:za3k.com ~all"
```
You should also be verifying this on your end as part of combating spam, but as far as outgoing mail all you need to do is add a TXT record to your DNS record.
Set your rDNS (reverse DNS) if it’s not already. This should point at the same hostname reported by Postfix during SMTP. This will be handled by whoever assigns your IP address (in my case, my hosting provider).
Test your spam reputability using https://www.mail-tester.com or https://www.port25.com/support/authentication-center/email-verification. You can monitor if you’re on any blacklists at http://mxtoolbox.com/blacklists.aspx.
Set up DMARC. DMARC declares your policy around DKIM being mandatory. You can set up whatever policy you like. Mine is
```
_dmarc 10800 IN TXT "v=DMARC1;p=reject;aspf=s;adkim=s;pct=100;rua=mailto:postmaster@za3k.com"
```

My sources writing this:

https://wiki.archlinux.org/index.php/Postfix and http://www.postfix.org/postconf.5.html on Postfix
https://wiki.archlinux.org/index.php/Dovecot on Dovecot
https://help.ubuntu.com/community/PostfixVirtualMailBoxClamSmtpHowto on SASL auth integration between Dovecot and Postfix
https://www.digitalocean.com/community/tutorials/how-to-set-up-a-postfix-email-server-with-dovecot-dynamic-maildirs-and-lmtp and https://www.digitalocean.com/community/tutorials/how-to-configure-a-mail-server-using-postfix-dovecot-mysql-and-spamassassin on SQL integration in Dovecot
https://scaron.info/blog/debian-mail-spf-dkim.html for setting up SPF, DKIM, and rDNS

Takeaways

You can set up store-and-forward mail servers, so if your mail server goes down, you don’t lose all the email for that period. It’s generally a free thing.
Postfix’s configuration files were badly designed and crufty, so you might pick a different SMTP server.
Email was REALLY not designed to do authentication, which is why proving you’re not a spammer is so difficult. This would all be trivial with decent crypto baked in (or really, almost any backwards-incompatible change)
The option to specify a SQL query as a configuration file option is wonderful. Thanks, Dovecot.
Overall, although it was a lot of work, I do feel like it was worth it to run my own email server.

Sometimes I have a bunch of dependencies. Say, UI components that need other UI components to be loaded. I’d really just like to have everything declare dependencies and magically everything is loaded in the right order. It turns out that if use “require” type files this isn’t bad (google “dependency injection”), but for anything other than code loading you’re a bit lost. I did find dependency-graph, but this requires the full list of components to run. I wanted a version would you could add components whenever you wanted–an online framework.

My take is here: https://github.com/vanceza/dependencies-online

It has no requirements, and is available on npm as dependencies-online.

I want to talk about three different mindsets for time management and what constitutes “success”. In all three, I’ll talk about an example problem, let’s say “writing term papers” for a student who finds this difficult, and see how each might approach the goal. My goal is not to promote one of these as best, but rather to illustrate that all three exist. Each may be favored by one person or another, or by a single person depending on the situation. I hope that by describing them, I can help people understand one another’s motivations, so as to facilitate communication. The first mindset I will call optimization. The optimizer tries to spend their time gaining resources. For example, they might work to get money. They also attempt to increase the rate at which they can gain resources. Some optimizers even try to increase the rate at which they can e.g. network or learn skills. The intuition here is that most goods are somehow fungible, and that you should try to get as many fungible goods as possible. Example of term papers: An optimizer might try to learn to write term papers, or get faster and faster at writing papers. If they got good at writing term papers, they might try to write even more (for example, taking classes heavy on papers) to take advantage of these skills. Heuristics:

Get nowhere faster: Get where you’re going faster, even if you don’t have a specific goal in mind
Level up: It’s always good to learn things and develop skills
Experiment: Never be content with the status quo. (Choose explore/exploit explicitly and dovetail, for principled experimenters)
Accumulate multipliers / Seek good trade routes: Be able to buy all things with money, learning, whatever you have most of. Try to maximize the rate.
Get money: Anywhere you have a resource coming in, increase the income rate, or trade rate you’re getting it for.

Failure modes:

Butterfly: Ends up unfocused and with no immediate or terminal goals.

The second mindset I will call satisficing. The satisficer is goal-oriented, and tries to stay focus on the goal at all times. Traditional goal-based time management like GTD encourages satisficing. Example of term papers: Their specific goal might be “Get an A on all papers until I graduate”. If they can write well enough to get an A, they probably would not learn to write better. If they can’t get an A, their goal might be better served by learning to write to rubrics. The question of learning to write better vs spending more time editing on each paper would be decided by the amount of time each would take for the actual number of term papers the student expected to write for their immediate goal. Heuristics:

Get to the goal as quickly as possible
No lost purposes: One of the strengths of satisficing is to avoid “rabbit holes” which don’t contribute to any end goal
Munchkin: Think about what will actually contribute to the goal, as well as what won’t.

Failure modes:

Tunnel vision / inflexibility: Not contributing to the immediate goal is not the same as being useless. Can lose sight of the big picture (supergoal or unrelated terminal goals) and ways to work toward that other than via the immediate goal.
EA Bot Syndrome / Don’t smell the flowers: Excessive goal focus can lose sight of human involvement, and end up giving the satisficer low life quality.

The last mindset I will call minimization. The minimizer will try to minimize the amount of resources spent on a particular task or problem. They will especially try to avoid any indefinitely-big cost. Example of term papers: The student might try to spend the minimum amount of time possible on the paper to get an A. If they knew they were going to write several papers, they might study how to write papers faster, so as to spend less total time–even if they expected the time learning to write faster would be greater than the time to write the immediate papers (“but what if there are more papers someday”). The eventual state of the minimizer will probably be to have set things up to avoid term papers, or write them very quickly. Heuristics:

Automate / Offload for free: Spend only a finite amount of time on something
Eliminate recurring costs
Asymptotic improvement: Any finite number of mistakes is okay, as long as you end up optimal eventually
Deal with it permanently: If you’re not closing things and returning to them indefinitely, you’re spending infinite time and attention on it
Timebox: Only spend a finite amount of time on something, and then consider it resolved forever

Failure modes:

Bad at goals: Minimization deals with recurring activities, but can fail to offer any positive motivation for one-time end goals
Negative mindset: Heavy focus on mistakes and resource use

Here’s how you make roasted chickpeas (aka garbanzo beans):

Set the oven to 400F.
Drain and empty a can or so of chickpeas into a collander and wash them
Dry the chickpeas (this is the hard step). I use paper towels, but I haven’t figured out a way to not use a billion of them.
Put them in a short pan in the oven and cover them in olive oil. Toss them some with your hands to get them coated.
Cook for 20-30 minutes, shaking the pan so everything turns every 10 minutes. I like them crispy so I do 30 minutes.
Take them out and transfer them to a bowl. Add spices. I like salt, garlic powder, and pepper.

A small font I designed. I’m pretty proud of it.

Install jq and youtube-dl

Get a list of the last 100 URLs:

curl https://api.twitch.tv/kraken/channels/${TWITCH_USER}/videos?broadcasts=true&limit=100 | 
  jq -r '.videos[].url' > past_broadcasts.txt

Save them locally:

youtube-dl -a past_broadcasts.txt -o "%(upload_date)s.%(title)s.%(id)s.%(ext)s"

Did it. youtube-dl is smart enough to avoid re-downloading videos it already has, so as long as you run this often enough (I do daily), you should avoid losing videos before they’re deleted.

Thanks jrayhawk for the API info.

Install Redis: https://www.digitalocean.com/community/tutorials/how-to-install-and-use-redis
Install Webdis
Start a minecraft server with computercraft. You will need to have the http API enabled, which is the default.
Put down a turtle I recommend a turtle with a crafting square and a pickaxe. I also recommend giving it a label. If you’re not trying the turtle replication challenge, either disable fuel or get a fair bit of starting fuel. Write down the computer’s id.
Put down a chunk loader, if you’re in a modpack that has them, or DON’T log out. Computers and turtles can’t operate unless the chunks are loaded. If you’re putting down a chunkloader, I surrounded them with bedrock for foolproofing.

Open the turtle and download the following script, changing “redis.example.com” to your own redis instance: pastebin get 8FjggG9w startup
After you have the script saved as ‘startup’, run it or reboot the computer, and it should start listening for instructions.

redis = "http://redis.example.com" 
queue = "sshbot" .. os.getComputerID()
return_queue = queue .. "_return"
print("Remote webdis queues on icyego: " .. queue .. " and " .. return_queue)
print("Receiving remote commands.")

function exec(str)
  print("Running: " .. str)
  f = fs.open("tmp", "w")
  f.write(str)
  f.close()
  p = loadfile("tmp")
 status, err = pcall(function () p = loadfile("tmp"); return p() end)
  if status then
    status, ret = pcall(function() return textutils.serialize(err) end)
    if status then
      result = ret
    else
      result = ""
    end
  else
    result = "Error: " .. err
  end
  print(result)
  return result
end

print("Now receiving remote commands.")
while true do
  handle = http.get(redis .. "/BRPOP/" .. queue .. "/5.txt")
  if (handle and handle.getResponseCode() == 200) then 
    str = handle.readAll()
    handle.close()
    str = string.sub(str, string.len(queue) + 1)
    result = exec(str)
    if string.find(result, "Error: ") then
      result2 = exec("return " .. str)
      if string.find(result2, "Error: ") then a=0 else result=result2 end
    end
    http.post(redis, "LPUSH/" .. return_queue .. "/" .. result)
  end
end

On your local machine, save the following, again replacing “redis.example.com”:

#!/bin/bash
function send() {
  curl -s -X POST -d "LPUSH/sshbot${1}/${2}" "http://redis.example.com" >/dev/null

}

function get() {
  curl -s -X GET "http://redis.example.com/BRPOP/sshbot${1}_return/20.json" | jq .BRPOP[1]
}

if [ $# -ne 1 ]; then
  echo "Usage: rlwrap ./sshbot <COMPUTER_ID>"
  exit 1
fi
ID=$1

while read LINE; do
  send ${ID} "$LINE"
  get ${ID}
done

Run: rlwrap ./sshbot , where is the turtle’s ID. You should be able to send commands to the computer now.

Install packages talk, talkd, inetutils-inetd
Make sure users have set ‘mesg y’
‘talk user@localhost’. Despite documentation, ‘talk user’ will not work.

So have you ever used a web printer and it was great?

…

Yeah, me neither. It’s probably possible on windows, but try to add more than one OS to the network and it’s horrible. And actually printing is a major pain in Linux anyway. Theoretically ‘lp’ and the like have no problem with remote printers, but I wanted something I understood. So today I’m going to post my setup I use instead.

I have a computer physically connected to the printer. Let’s call it ‘printserver’. On that server there is a folder, /printme, which is constantly monitored by inode. Any file added to that directory is printed.

Suppose I downloaded cutecats.pdf and I want to print it. Then I run:

scp cutecats.pdf printserver:/printme

And voila, the cute cats get printed.

Here’s the setup for the server:

Get the printer to work. This is the hard step.
Make a directory /printme. Add any missing users, add a new group called ‘print’ and add everyone who needs to print to that, etc.

Set up /printme to be a tmpfs with the sticky bit set. (So we don’t fill up the hard drive)

/etc/fstab
tmpfs           /printme        tmpfs   rw,nodev,nosuid,noexec,uid=nobody,gid=print,mode=1770,size=1G  0       0

Install incron and add this to the incrontab (of user ‘print’ or ‘sudo’):
```
# incrontab -l
/printme IN_CLOSE_WRITE,IN_MOVED_TO lp $@/$#
```
Note that this will preserve files after they’re printed, because my server is low-volume enough I don’t need to care.

In space no one car hear “whoosh” sounds but action potato is so cool you still can mostly.