Your mail will go to Spam too with Maddy, this is not about the MTA.

This is not (always) true.

Maddy comes with a lot of the 'good stuff' like DKIM and DMARC ready to be used, whereas setting that up yourself with Postfix was a pain (from my experience a few years ago) and probably meant you didn't bother.

I've had better deliverability, especially to Gmail and Microsoft, since using Maddy, probably for this reason.

Now, of course, I must admit that self-hosted e-mail seems like it will always inevitably run into a blocklist problem, usually because of IP neighbours. But personally, I've managed to avoid that in the last year that I've been running Maddy -- main exception is I had to fill in Microsoft's silly form to get unblocked at the start (but I've had to do this for every deployment I've ever done).

I dunno, I'm pretty shit when it comes to Unix systems and programming in general (spent 6 hours fighting a postcss and tailwind config file and lost) but I got dkim and dmarc set up on my own server for my own mail in maybe 30 minutes?

So you agree is not about the software but you did not implement best practices with Postfix. Also, whenever you give an address to someone else, family or friend, risk of going to spam increases. Links or src in emails without https, higher risk, Does Maudy prevent you from doing that?

If one software makes best practices easy to implement and another makes them hard, it still is about software. Not entirely, but it's important.

It is definitely possible to configure something like rspamd to scan outbound messages.

And your sales people will call this a misconfiguration serious enough to fire the sysadmin who did this. The problem is that, typically, nobody trains or supervises the filter applied to the outgoing mail, and nobody is warned that it did, in fact, reject an email, and there is no manual override if the filter is, in fact, wrong. And spammers (from the viewpoint of a particular organization) don't exist and can't exist inside the organization, so any rejected outgoing email is, by definition, a false positive.

Well, viruses are the exception.

One "weird trick" I discovered with deliverability was that plaintext email from my server was rejected, but HTML email from it was accepted.

For anyone with deliverability issues and mail going to spam, it might be worth trying this out (after you have sorted all the usual points like DKIM, SPF, PTR record etc.) - this obviously only helps if your IP is clean. I was surprised though, and only discovered it by accident.

I had good deliverability to Google hosted mail, but not to Microsoft 365. I even had issues with deliverability doing 365 to 365, until I sent HTML mail.

What a world we live in, I always would thought the opposite. Probably it was like that before, but now maybe spammers send more plain text e-mails to appear legit. Never ending game of cat and mouse.

My guess (pure speculation) is that more spam is auto-generated via email libraries and similar, and that is producing plaintext mail to try to simplify what it's getting through the filters.

On the other hand, most human-generated mail is HTML-based, as it's written in a web-client or desktop/mobile "app client", that is using HTML to let people do bold and put pictures into their email.

I was as surprised as you when I figured this out!

As someone who has run his own mail servers since 2000, and still gets his email delivered to people's spam folders (even when I've emailed them in the past, or simply replied to their original message), this is very interesting, as I always enable plaintext outbound email.

I'll have to give this a try. Thanks for the tip!

Your symptoms sound exactly like the issues I was having - everything was "right" (in theory) - emails were leaving on the right IPv4, that had the correct PTR, etc etc. I just couldn't figure out what was wrong.

It was the same issue - even replies to existing threads had the issue. What got me looking into it more closely was that I had no issues delivering to G Suite tenants; it was only issues with 365 tenants.

I wonder if there's a non-trivial correlation between people choosing to genuinely self-host their own mail, and those who use plaintext outbound mail by default (and know what that means).

That's odd. I know that Spamassassin actually penalizes HTML and HTML only email.

Are you sure there is not some sort of parasitic Bayes filtering happening here?

It's very odd. I don't know how Microsoft implements their own server-side spam filtering, but the issue was coming from their filters. No issues with any other host. G Suite accounts received plaintext mail fine.

I never found much by way of technical detail on 365 deliverability, beyond the usual DKIM/SPF/DMARC/PTR record guidance, and forms for getting yourself removed from the whitelist.

What's especially strange is that this issue applied even when sending mail as a 365 subscriber. Once there was HTML content present in the email, deliverability (to other separate 365 tenants) improved!

Very odd...

My mail goes to spam using GSuite for my Domain for Work Gmail! (or whatever it's called these days)

Where do I even turn? Did my startup's domain get blacklisted before I've even launched?

I also hated the headache of Postfix so I switched to OpenSMTPD on OpenBSD and it is so so much nicer. It feels much more set-and-forget than any other time I've tried running mail servers on Linux.

> Lately I've been seeing an uptick of these apps making self-hosted email easier. Mail in a box, Maddy, etc. I will unconditionally support anything that makes doing this easier.

See also ISPmail (which uses Debian):

* https://workaround.org/ispmail

I had great fun setting up Postfix and Dovecot recently.

However, I was very specific with always going for the minimum most simple solution. I do not have a dedicated database for instance, but use just mbox and POP3.

And I did not even bother much with anti-spam, I chose a super easy filter in Thunderbird instead, sending everything with 'unknown' in Received: header to the Thrash folder.

Getting accepted by others was also not much of an issue (but lots of details of course, PTR, SPF, DKIM, DMARC).

More details here if anyone is interested:

http://bef.no/DitchingWindowsAndAWS/

> And I did not even bother much with anti-spam, I chose a super easy filter in Thunderbird instead, sending everything with 'unknown' in Received: header to the Thrash folder.

I realise that everyone hates spammers, but having a "thrash" folder is a bit much, isn't it? ;)

I haven't tried Maddy (I'm sure it's quite good), but I do use Cloudron right now and it's a breeze to manage email in it.

How would you compare it to a non-trivial Apache or NGINX install? What about configuring an environment for Rails or Docker or…?