[Discuss] Running a mail server, or not

[Rich Braun]

Derek Martin <invalid at pizzashack.org> raised a couple more interesting points:
> The fact is your
> e-mail is already being consumed by the great government surveillance
> machine regardless, since both incoming and outgoing mail has to
> traverse multiple ISP backbones 

Not mine, at least not in clear-text. Backbone providers only see encrypted streams between my email server and my service providers' systems located in France and Canada. I'm not aware of any government surveillance that siphons off regular users' encrypted (SSL) transmissions for decryption later: there's just too much of that data for today's technology except for targeted cases where a government has reason to look at a specific data stream. (Remember, every SSL website prefixed https: uses the same type of encryption that my email server does.)

In order to assemble all of my email traffic as plaintext, an intruder would have to specifically target my mail server (some type of phishing that would get a password /and/ a backdoor through my multi-factor auth), or figure out the identities of my correspondents and go after their less-secure systems (a large fraction of whom are on gmail or yahoo).

> I do not use IMAP--I log into my server remotely
> over SSH to read my mail with Mutt. I've found over the years that
> this was the simplest way to ensure I could actually get at my mail
> without being blocked by firewalls

One place where I worked blocked port 22 but not 443. That makes the case for webmail servers like Squirrelmail, Rainloop, Roundcube and/or a couple of others. But I do still have a few things that I like to run from a command line: for those, or for web UIs that I don't want exposed on the Internet, I use a remote-desktop access tool called Guacamole, which shares screens or ssh sessions through an https connection.

> And to be honest, now that I'm getting older, I've started to think
> about what happens if I should die. Frankly, no one will be able to
> figure out my hosted server details

Estate-planning is part of why I overhauled my systems to the current state-of-the-art. Living in the middle of the tech universe (everything seems to have moved three thousand miles from 02139 to 94107 in the past 10 years), I'm bombarded with memes related to containerization or cloud-sync every day. So, my email systems could be figured out by today's 20-something techies who happen to be friends of whichever family member needs access.

Create a doc explaining your tools, make as much of it public as you can (another aspect of my estate plan is the README at top of my main github repo, any of you can go there to see and/or decipher what I've done with my systems--so you're all enlisted to help the executor of my estate--hopefully many decades from now but before the USA falls into the chaos we're all worried about). Create a USB flash-drive with your system-startup credentials and more-specific instructions about private details of your setup, print those out and stick them in a safe-deposit box whose key is held by whomever is designated to go pick up your body after the proverbial hit-by-a-bus incident.

Arguably, having most of your private data (pictures, videos, writing, art, whatever) on your own private systems makes it more-accessible to your heirs: mine is all in one place, accessible by one set of credentials. Most people have stuff scattered across many cloud-based services, with different credentials--and a lot of it will get entirely forgotten as our memories fade over the decades.

By chance, I happened not to have any itch to ever run for President. So my private email server never became particularly controversial. ;-)

> I do also maintain and use a gmail address, and over time, I've been
> increasingly relying on that for convenience.

The one reason I too have a gmail address is as a backup for the handful of (usually older) services that refuse to send to or receive from my personal email address.

> to stick with my current scheme, I'd have
> to create an e-mail for them on the fly, and find a way to actually
> create it before they're going to use it.

Containerizing makes it easy to script generation of domains/aliases for deployment (to your postfix, spamassassin, dovecot, email client, anything else in your tech-stack) in seconds.

-rich