This isn’t universally true, although your broader point (it could change at any time!) is a risk.

As a data point, I have run my own email servers for years. Both professionally (couple million transactional emails per year) and for my personal.

If you follow best practices (which is a pain to setup!), deliver ability is generally on par with other providers, unless you got unlucky and got a bad IP.

This is my experience too: there are a couple providers that are more difficult (Verizon, iirc), but I’ve never found a personal email server much of a hassle to maintain.

I run my own email server, and have for many years. More recently, Verizon/Aol/Yahoo have been a pain in the butt for our larger email lists, even though we run a very clean operation. Still, our deliverability is, on the whole, better than through the much more expensive providers, such as Mailchimp.

I second that. Around 10 years ago it did consumed some time to set up everything, but since then I didn't touch it at all.

Much better and more flexible than any proprietary mail system.

I've recently given up on running my own email server and changed to postmark. The deliverability has gone up, but most importantly I spend less time managing it.

Postmark would cost us around $10k a year. I spend maybe 10 to 20 hours in a year dealing with the server and deliverability issues. It's not clear to me that Postmark could improve our deliverability much either, as it is already pretty good. How much did it improve for you?

Mailinabox is a great package that makes the best practices very easy. I always recommend it or similar packages.

This looks awesome, thanks.

If you want clean IPs for self-hosting a mail server, you can use a service we created called Hoppy Network:

https://hoppy.network

Our IP addresses are not on any blacklists, and we don't block SMTP or mail ports.

for now rubs hands

I also just found out about a service called Tailscale (maybe 30 mins ago). It's not a field I have investigated but both your services and theirs sounded vaguely useful for some problems I'm having. Are they somewhat similar? It sounds like I could solve the same problems in perhaps two different ways.

I think they are pretty different, just both using the same underlying protocol.

Tailscale (which we have been using to great effect over the last year or so) provides effectively a 'LAN' where the devices can be anywhere. It provides you with a second interface in the 100.x.x.x range. You can trust that any traffic from that range has been authenticated to connect to your Tailscale network.

Hoppy (as I understand it, I've never used it) is more solving the issue of getting a static IP assigned, with the advantage of it following your device if it moves between networks. You get a second interface that is publicly visible, but can't trust traffic coming in from it.

that's very interesting, thanks!

This is a recurrent FUD dropping. I’ve run my own for about 15 years and it works. Occasional mails get sent to spam by Gmail, but that can happen if you use a big provider, too. Take an hour to set up DKIM, etc., and you should be OK.

> Occasional mails get sent to spam by Gmail, but that can happen if you use a big provider, too.

In fact it'll happen even when you send from gmail to gmail itself!

My deliverability from my own email server to gmail feels better than from my company account (on gmail) to gmail users.

I just switched my long running SMTP to a new IP. I've got SPF, DKIM and DMARC all dialed in. I'm not getting tagged as spam. Even when I add new domains to this host for sending - still gets through to G, Y and MS based services.

As SPF, DKIM and DMARC have become more widely implemented, is IP reputation less of a factor than it used to be in spam detection algorithms? This is just my speculation, but it seems plausible.

Why would it be less of a factor? These things don't prevent spam. I can create a proper mail server with all these and send billion messages from it wherever I like. These technologies prevent various kinds of forgeries.

Quick look at my archive, of 37k spams only 7k fails SPF. (I accept all email no matter what.) So it's some signal, but not nearly enough.

SPF will pass if there is no SPF record in the domain. Which most probably is the case for most.

Look for DKIM. Either it's signed and pass. Or spam if DKIM is missing. You can filter a lot more this way.

dkim=pass is on 5k spam messages, missing dkim is on 95% of HAM email I receive. So that's a no go and not a great signal anyway.

> These technologies prevent various kinds of forgeries.

Preventing forgeries is what allows the trust to be based on the domain/address rather than the IP.

I don't know. But a) I'm low volume and no list services and b) I checked my IPs against the BL before setting up the VMs. I still believe the terror-tales of trying to get an IP off them. My own spam filters still use them.

SPF, DKIM and DMARC are more about anti spoofing than spam prevention.

Not true. With some effort, it's possible to set up a pre-made email server script like Mail-in-a-Box or Mailcow and make sure its reputation is clean enough to send and receive email from the big providers like Outlook, Gmail and Yahoo.

It's much easier these days. The ecosystem has gotten more complicated, but the community has put in a lot of effort into automating that complexity away.

The only “common” email provider that has given me problems is Zoho. They’re annoying and do not respond on their mailer daemon email or whatever. I tell people we can’t correspond if they use Zoho and no one has had a problem with that yet.

I want to throw my hat in here because there are plenty of the "works for me" crowd.

I've run a personal mail server for the last 5 or so years and I have SPF, DKIM, DMARC setup correctly, get 10/10 for mail-tester.com and get graded "not spam" for all the other spam tester sites. Spam Assassin loves me and I have a .com mail server.

I can land more or less into personal emails (occasionally I get filtered into spam), but when it comes to corporate emails I pretty much get /dev/nulled each time and can't even land into a spam folder. I can't give out my personal email as a point of contact and have to use my gmail or protonmail instead.

I've all but given up on owning my own email because the protocol has effectively made it impossible to reliably land mail in a mailbox if you don't have a certain amount of outbound mail traffic.

did you send to gmail/hotmail/icloud?

hotmail and icloud are very sensitive and if IP isn't reach a certain reputation(mail volume+age) it's easiser to land on spam even SPF/DKIM/DMARC are passed. But if the recipient interact with sender(mark as non spam, send a reply) then moving forward email send from that won't be flag as spam anymore. So their spam filtering is kind of weak and they used previous interaction as a signal.

gmail, on other hand is rock solid. They have their own false positive obviously but they are rock solid in term of not relying on IP reputation(or less). They are smarte and righfully land your legitimate email on inbox even an email come from a spammy IP.

I've not tried icloud, but for personal email addresses that I have tried (gmail, outlook, hotmail, protonmail) I've not had many problems landing in inboxes. For corporate mailservers I almost always have trouble even getting beyond the mail filtering solution. Microsoft EOP (which seems to be the most ubiquitous) will basically blackhole any mail I send without telling me.

You still can. The main and most important thing is to have a clean IP. So no major cloud provider who reuse IPs.