Outlook.com is silently discarding email messages (2015)
blog.paranoidpenguin.net
blog.paranoidpenguin.net
The unfortunate thing is that we also find the demographic of customers here in the UK still using btinternet.com email closely aligns with customers who tend to be more problematic…
(My parents have a btinternet.com email address)
Yes, the demographic of a btinternet email user.
Maybe this is just wishful thinking, but if the more problematic customers are losing you money, then isn't it a valid business case to stop doing business with them given that you can identify them in advance? Wishful thinking because it'd be really nice to create a feedback loop such that users of email providers start to understand their reputation, which would make provider reputation for deliverability matter, which might improve their behaviour. But most businesses probably aren't in a position to turn away business like that.
They actually stopped me hosting my own email because I couldn't be bothered to deal with this shit any more.
> IMAP< 72 NO [LIMIT] UID COPY Rate limit hit.
> \* IMAP error on imap.aol.com: UID COPY error
You'd think you'd be able to log in to your Exchange dashboard and just look at them. They hide it from you and only show you abbreviated logs. You can go through support, and if you pay enough money you'll get an cryptic answer within a week (if you're really lucky the support person might show you the real logs).
Pain in the ass when you're trying to figure out why your users aren't getting email from someone.
You can pry the mail logs from my cold dead hands.
You'll need everything set up and E5 licenses if I remember correctly.
Tons and tons of good info though.
With Office 365 I have far less issues. So Microsoft can do it, if they want to.
$ dig mx outlook.com
;; ANSWER SECTION:
outlook.com. 300 IN MX 5 outlook-com.olc.protection.outlook.com.
$ dig mx uniri.hr
;; ANSWER SECTION:
uniri.hr. 86251 IN MX 0 uniri-hr.mail.protection.outlook.com.The free version has probably a lot of features turned off.
If it was proven in court that they were silently discarding correspondence of paying customers because they were over confident in their spam filtering, everyone would jump ship in a heartbeat.
Mail is a very resilient protocol, this isn't UDP, this is the result of a deliberate decision being taken.
If storing vaguely suspected spam is intolerably expensive then maybe just keep a list of discarded messages (sender, title and date). At least users can see what's happening.
The server rejects the message, to get rid of spammers. After a few rejected messages, they usually stop delivering SPAM for quite some time (it's unnecessary work to do, and a higher risk to land on blacklists).
Mail delivery is an anarchy and I’m impressed the problem of spam has been solved to the degree we currently see. Yes it’s much more centralized than before, and no that isn’t a good thing, but from a customer’s experience point of view, I couldn’t go back to the old days of manually training your email client’s spam filter and whatnot.
Certainly not normal, but done sometimes, yes.
> and one of the reasons it can be painful to manage your own SMTP server, right?
It's part of the job, but it's more likely that the big providers are the least of your problems. They can be reached and you're usually not alone with your problems.
It's the small providers who have terrible filtering, who use low-quality shit-lists like UCEProtect/Backscatter or their own "heuristics", that are the most painful.
Why can't the big ones just check SPF and DKIM etc and let us responsible self-hosters do our thing
I really don't think Microsoft would ever touch that pile of shit, most likely your spammy neighbour got your neighbourhood blacklisted at both places simultaneously.
> Why can't the big ones just check SPF and DKIM etc and let us responsible self-hosters do our thing
Most of the spammers have valid SPF and DKIM. Quite a few have entire subnets of IP addresses to spew spam from. That's why a random IP and just those two aren't sufficient to just start blasting.
What porbelm means is that when SPF and DKIM are valid, reputation check should be based on the sending domain, not on the sending IP.
This way spammers still get thwarted (because of their domains with zero or low reputation) and legit senders can send from any IP (since their domain CAN build good reputation)
Besides, it's also easier to track reputation of ~360M domains than billion of IPv6 and IPv4 ranges.
Unfortunately many legitimate sending domains are newly registered, so have no reputation. It's not possible to rely on age either: if requiring domains to be aged becomes commonplace, they're cheap enough that spammers would just register them and wait for a bit, maintaining a pipeline of them.
So it becomes necessary to consider the reputation of whoever is enabling the sending of the email. The only way to do that is by IP.
If a domain uses DKIM, has a good reputation but comes from a bad IP, then sure, you might trust it. I don't think that'll help much in practice though.
Consider this: every legit IP address once started with the problem of having zero reputation.
How you build reputation from zero is simple: you start by being allowed to send only small amount of daily emails (rest goes to spam or is outright rejected by the recipient SMTP server). Then adjust reputation based on the usual factors (observe human recipients flagging emails as spam/not spam, watch for keywords like Viagra, etc, all the exact same stuff email providers already do). You gain or loose reputation over time, which affects how much you can send: a lot, or nothing.
According to your previous post, I shouldn't do that for DKIM/SPF -valid emails by IP, but I do. I hope this explains I do, and why purely domain-based reputation doesn't work.
So, in your example, blocking by IP isn't necessarily superior to blocking by domain (and vice versa.)
Once again: it's possible to have the tools and infrastructure manage domain reputation and do a job as good as (and arguably better than) IP reputation. Also to consider: IPs can be acquired by spammers for nearly $0 (eg: destroy+recreate VPS to get a new IP for free), whereas domain names have a fixed cost of ~$10 USD per year per domain. So if you have a system that manages domain reputation and has 1 million domain in the blacklist, you know for a fact that you are burning ~$10M/year of spammers' money.
Recent trend I'm seeing is creating malicious Google and Microsoft account or hacking websites and using those to spew spam.
I self-host and frequently block entire ISPs. It's very common to get the same mail for days on end from a different IP, and often a different domain each time. Some months DO accounts for over 50% of my spam, for example.
I'm sure there's more graceful ways to handle it if I was willing to put far too much of my own time into fighting a losing battle. but adding a REJECT line for a whole netblock takes seconds.
I don't begrudge the big hosts for treating whole ISPs as cesspits as I do the very same thing.
From what I see, it's just a few players exchanging mail between themselves and discarding a high percentage of traffic.
Yes. Silently discarded e-mail is the worst thing that can happen because you can't detected it as a sender. Your message reaching the spam folder is bad, but not as bad as getting silently discarded.
The best among the bad options is probably getting rejected with the SMTP return code 550 along with instructions how to delist in the return message.
Google's Postmaster tools all show "No data to display at this time. Please come back later. Postmaster Tools requires that your domain satisfies certain conditions before data is visible for this chart."
Microsoft's Safe Network Data Service shows *"No data for specified IPs on this date" for every day.
There is almost never any valid reason to block an entire ip address. Single user accounts (e-mail addresses), maybe, if compromised.
E-mail, as a standard, is still pretty straight forward. Ideally you should be able to send e-mail directly from your laptop IP, if approved to send e-mail on behalf of your e-mail account / host name. Decentralization is important.
I'm sorry but that's just very naive.
It would be fair(er) to explain why.
As things currently stand, the only "expensive" resource that is hard to acquire or fake somehow is source IPs with good reputation. Fair? I'm not saying it is. It's just effective.
A domain is relatively cheap. Accepting some 20 messages before you are sure it's spammy is also cheap. You will have to filter out the malicious domain registers, but that's a quite immutable set, as becoming a register is a bureaucratic process with a non-trivial cost.
One might think that would be the case eventually, but at least in the case of universities insisting on using Office365, it really doesn't seem to be. In academia, where emails to and from university domains are the norm, Office365 seems to frequently treat with extreme skepticism odd domains, like those with .edu or .org TLDs, or prominent universities on ccTLDs in Europe, assuming that most legitimate mail will come from inside an organization. I've seen it send directly to spam emails about papers, peer review, talks, emails from easychair.org, emails from major journals. I know of a case where it has sent emails from Nature's submission system to spam. I've seen it send emails from frequent collaborators to spam. Classifying emails as not spam seems to do little to nothing. Meanwhile, emails from major providers' email services, like gmail and outlook, often seem to get through, despite mostly being spam, and in some cases, rather obvious spam.
It seems likely that people have lost publications and talks as a result: the emails being rejected and spam-filtered are important emails. Yet as much as everyone seems to complain—and these are complaints I have heard from many people, at many universities—Microsoft seems to have continued success in selling Office365 to university IT services, and pushing everyone to use it.
There are likely some silently discarded e-mails as well, but you can't know for sure because they weren't even delivered.
The trouble is, how many users are sufficient? Patch Tuesdays have their reputation for a reason, and if your issue isn't one shared by millions of people, it's unlikely a fix will come from Microsoft. Meanwhile email continues to get lost, and Microsoft ignores the issue.
What are you aiming to accomplish by blocking a cloud services provider and not a particular sender domain?
When I gave up on SES, I was getting annoyed by a commercial list that had no unsubscribe (it did, but went to a domain that didn't exist). So I had a grep through to see what I'd be missing, and it was 100% worth missing. I don't think I've seen any other provider that'd be 100% - usually when they get that bad it's just here-today-gone-tomorrow hosts.
The impression I get is that providers like mailchimp where they're actually packaging it as a service, so they handle unsubscribe, campaigns, they have an abuse@ contact that actually works .. these services deliver content. Dumb pipes that only exist because the cloud provider's regular IP space has a trash reputation, tends to carry everything that gave the regular IP space that reputation in the first place. They're just weaponising the idea that no-one would be dumb enough to block amazon. I'm dumb enough.
Also self hosted, oldest domain probably about 15 years - dropping the above hasn't actually affected my as negatively as people getting gsuite wrong and failing dkim etc
The Outlook teams are weird. They all have their own special feedback mechanisms that are different from the rest of Office, and they for the most part ignore them.
Earlier this year, for some reason the SMTP servers were changed from smtp.live.com to smtp.office365.com, breaking a number of my workflows and integration with other tools. This is meaningless pain that served no point - you could just point the DNS records for smtp.live.com to the exact same servers smtp.office365.com points to. Combined with other things Microsoft has done (breaking decades of links to MSDN blogs and support pages without providing redirects) I have no faith in the stability of the product in the future.
Their web view generally sucks. It doesn't play nicely with the back button - selections are lost, search results are bypassed. It gets randomly stuck where it won't load or will load the wrong CSS for hours at a time on multiple of my machines.
The Android and iOS apps have largely not changed since they used to be Accompli. I can't name a new feature in the last decade.
Why, in the left nav bar, is there an icon with an envelope and a plus button that is "connect a GMail account"? That icon suggests "write a new mail", which is a function I would do quite often. I never want to connect a GMail account - much less have it take up a dedicated button that's always present in the UI.
Outlook on Android has its section buttons (mail, calendar, etc.) on the bottom. Mail for Windows has them on the bottom. Outlook 2016 and 2019 have them in the lower left. Outlook.com for many years had them in the lower left (if vertical). At some point they've moved to the upper-left, which is inconsistent with every other Microsoft-provided way that I check my mail and a regular source of frustration when I throw my mouse and eyes to the lower left corner and find nothing there.
I could go on and on (my favorite bug is that they removed the Send Feedback button that their docs refer to so I can't tell them any of this). I was an Outlook fan for several years, but I could have written most of this feedback in 2015 and nothing's changed; at this point I would encourage anyone still using it to just forward their mail to some other provider and be done with it.
The document attachments which come along with incoming email are still listed in "Documents" tab in Outlook even after they have been purged from secondary trash. The documents themselves aren't available (cannot be opened) but a list of all documents which have ever arrived (including JPGs/PNGs in signaturss) are still visible in that tab. I flagged this issue once but no resolution yet
It's a ridiculous and easily dismissible statement at first, but not so ridiculous if it becomes commonplace/familiar/canon (it worked for internet explorer!)