Mailtrain Newsletter #8 – Bundled MTA | Hacker News Reader

Mailtrain Newsletter #8 – Bundled MTA

mailtrain.org

81 pointsandris936 comments

I've been using mailtrain in a business context (not talking about buy in newsletters) for a while now and I purely love it! It's fast, reliable and the ui is perfectly organized.

What I still miss is simple Email functions like adding attachements and CC receipients. This sounds rediculous, but the missing CC has been keeping us off from using mailtrain in a lot of campaigns where simply other people where important to be informed about replies.

Still I suggest anyone trying out mailtrain as it is super simple to spin up. I love seeing it getting new releases, thumbs up!

Attachment support was added just today, see the related blog post: https://mailtrain.wordpress.com/2016/09/09/attachment-suppor...

I don't really get what you mean by the CC recipients. Could you create a new issue in the Mailtrain issue tracker in Github and describe what you would like to have https://github.com/andris9/mailtrain/issues

For the lazy, has anyone done any cost analysis?

I've used Sendy (self-hosted php, but non-free licence) before, and that is tied to just Amazon SES.. but this has a few options.

I also use Sendy: one with Amazon SES and one with a more dodgy local Postfix server. I am quite happy with it, but the feature missing in sendy is Segmentation. How does this compare?

Mailtrain allows you to create segments over exact or relative matches eg. "all users where signup date is after 7 days before today" or "users who have custom numeric field value between 1 and 10". There's not too much documentation about it but you an find some from the Wiki https://github.com/andris9/mailtrain/wiki/Relative-date-rang...

Do yourself a favor and read the Sendy source code.

Can you elaborate?

I read on multiple places that it's a total mess, probably he means that :p

Here's one example of what I saw.

Imagine you want to track every open and every click an email gets. Most of the email clients nowadays block that hidden image that is used to track openings. So, if a user clicks on a link, that obviously means that they opened the email and an email open should be tracked as well.

In order to do that, one might write a function that tracks opens and call that on the tracking pixel and whenever a link in clicked.

Well, that's not what happens on that software. When the, for instance, `linktracker.php` runs, it makes a HTTP request (think open('http://$HOST:$PORT/opentracker.php?uid=$UID&campaign=$campai...) ) to the `opentracker.php`, simulating what a browser would do, in order to avoid duplicating the code.

I have a question for Amazon SES users. Do you need to submit the "Request a Sending Limit Increase" form to send emails to domains you don't own? Can it still be free? I just want to send 3-4 emails a day from my vps and my home router.

Anyone with any experience with this product? I've got a lot of users who want to use Mailchimp, but we run a very restrictive SPF/dmarc and we will not include any 3rd parties.

This might be a good solution for self-hosting our own.

The biggest value Mailchimp provides is deliverability. They spend a lot of time working with ISPs to maintain the reputation of their IPs to make sure their mail ends up in the inbox and not the spam box.

You can't self host that.

The mailing list features themselves aren't much different than anyone else offers.

I would argue that they absolutely fail in that regard. The default behavior from the service seems to be "impersonate" the user - any corporation that runs an SPF has to choose to either allow (tolerate) or deny it.

Well yes that's how SPF was designed. Of course if you have an SPF record they will fail (and so will anyone else) if you don't allow them. That's kind of the point of SPF -- to prove to everyone else who you trust to deliver mail for you.

It's probably fair to say most customers want a white label solution so that's a completely reasonable default. If you're sending out emails from mailchimp on behalf of your company why wouldn't you want them to be from your company?

Don't you get "deliverability" from the email provider (i.e., SendGrid, SparkPost, Amazon SES)?

Yes but in Mailchimp's case they are the provider. They send the emails directly.

Sure, but this doesn't actually send the emails. That's what you plug in Mandrill, SendGrid, SES, or <SMTP-provider of choice> for.

MailChimp charge crazy money for large numbers of subscribers. If your business actually makes money on direct-marketing, that's fine, but their pricing stings if you want your marketing team to manage large lists with only occasional informational emails. And since MailChimp will give you their deliverability for 25,000 emails for $20 on Mandrill, it seems like a win-win for that use-case.

I, for one, can't wait to try this out over the weekend.

Mailchimp also has pay-as-you-go pricing where you pay per email for lists where you only email subscribers occasionally. I think its around 2c per email (depends on how much credit you purchase at once)

> hey spend a lot of time working with ISPs to maintain the reputation of their IPs to make sure their mail ends up in the inbox and not the spam box.

This is the first time I hear Mailchimp (or any other company) has some special deal or treatment from ISPs in regarding to their sending emails reputation.

This is like saying lawyers of company X spends lot of time working with judges to make sure they are winning their cases. OR like suggesting to host your website with GoDaddy because they spend lot of time working with Google so that your keywords will show up on their first SERP.

Nobody is in position to just convince ISPs to deliver your mail better than any other competitor, because that would be a special treatment that would cry for gov oversight, investigation and/or prosecution. And besides, it would be totally worthless in terms of oversight. So let's say you do have special privilege -- then some Mailchimp client gets hacked and their account used to spew spam and before its shut down there are millions of malware installed. What would Yahoo say to their customers?? Simply put, no ISP would want to waste their time giving anyone special treatment when it comes to delivery...

The truth is if you are properly configured with all belts and whistles and are using feedback loops and your IP does not have dark history, then eventually building up your reputation (continuing to send emails that are most likely not spam) you will end up at the same delivery levels as Mailchimp & co. Again, because noone at Yahoo sits down by a monitor and browse billion messages a day and decide which one came from Mailchimp (good) which one didnt (bad). But I don't blame Mailchimp that they don't want you to know about it...

One of my colleagues is currently running $25 DigitalOcean machine with Postfix and sends opted-in ~5MM messages per day with stable ~85% delivery rate. Custom scripts and no Mailchimp required.

> This is the first time I hear Mailchimp (or any other company) has some special deal or treatment from ISPs in regarding to their sending emails reputation.

If you use Gmail and PayPal you can see this in action. There will be a little key there. You have to be an approved sender to get the key. The key, besides being a visual indicator, guarantees delivery in the inbox.

Most people who admin big mail sites know each other. When their mail is bouncing, they call each other and ask for help. When they see their friend's big site getting a lot of spam warnings, they call their friend and say "hey this word in your email is triggering our spam filter, you should change it".

This is just how industry works.

I don't use Gmail at all and have only used PayPal perhaps twice in the last three years so I can't be certain but you're not referring to their visual indicators for encrypted e-mail [0], right? You're talking about something different?

[0]: https://gmail.googleblog.com/2016/02/making-email-safer-for-...

https://i.kinja-img.com/gawker-media/image/upload/s--9UDz03_...

Using Mailchimp alone does not give you the golden key icon, you have to be PayPal for that which is not the case for most people

That's probably fine for doing your own thing but if you let your users generate email or you screw up and send something that's flagged to hell the established relationships you don't have with the companies that provide spam filtering and/or email is going to hurt. If you're just sending out occasional newsletters or whatever the time to resolve that probably wouldn't hurt but if you're sending critical email on a frequent basis (like password resets) any delay could have a big impact.

I certainly wouldn't recommend to use your primary domain name for sending out large e-mail blasts that might get marked as spam (whether they are "legit" or not).

An alternative might be to set up a subdomain (e.g. marketing.example.com) for those outgoing e-mails, create your SPF RR for it so that only MailChimp's sending IP addresses are permitted, and let your users use their preferred tool (Mailchimp).

How advanced is the segmentation in mailtrain? Can it segment lists like (ListA + ListB) - ListC

We're currently using Interspire Email Marketer and the segmentation feature is quite limited.

503 Service Unavailable?

It was that damn Apache that could not handle any load. Killed it and exposed Node.js directly instead of proxying through mod_proxy, should be fine now

Does nodejs protect against slow requests (probably not) ?

Slow requests are not an issue as all connections are handled by the same process. If a connection gets idling or slow, then it does not take almost any resources (compared to apache where a separate process is spawned to handle that request)

> compared to apache where a separate process is spawned

That's not how Apache works. If your glibc is >= 2.3.2 (we're talking 2004 or newer) the default mode is epoll ("event" MPM), otherwise you will get a threadpool ("worker" MPM), and failing that you will get a processpool ("prefork" MPM).

It will never spawn a process to handle a request, unless you forcibly enable prefork mode with a pool size of 1.

Sure, I agree that my knowledge about Apache internals might be severely outdated. And I can't really blame Apache either because it's a very small box, in a larger one it might have quickly stabilized

In my experience this is due to an Apache default configuration that probably seemed reasonable when introduced but in practice can really screw you. Basically it caches and serves only the 503 for something like 3s to prevent the application it's proxying from getting kicked while it's down/flapping/starting. So the end result is that even if your app is only occasionally returning 503s for a small subset of users everyone gets to enjoy them for several seconds every time one is encountered. Fun stuff.

It's a really small box (512mb) that runs everyting and on a normal day it does not have any issues but when the requests from HN started coming in Apache used up all the RAM. I stopped Apache and started Mailtrain from command line in a screen session to listen directly on the web ports (so once it's out from the HN frontpage I can reboot the VPS and everything gets back to normal settings). So far it has been running smoothly with the server RAM usage in a constant 50%, no fluctuations https://cloudup.com/cokRbDpMIuw

Yeah that's a pretty tight fit. Can't say I have much experience doing that all on one small box outside of local dev stuff.

Reply on news.ycombinator.com