Bug #1268

[proton] Yahoo mail servers sometimes reject our emails

lukeshu - about 6 years ago - . Updated about 6 years ago.

% Done:



I recently `git send-email`ed a patchset and it Cc'd (since I made the commits with that email, and didn't turn off the auto-Cc feature). forwards to , which is on Yahoo servers.

In a set of 2 patches, the first got through, and the second bounced with

<> (expanded from <>): host[] said: 553 5.3.0 alph148
    DNSBL:ATTRBL 521< >_is_blocked.For assistance forward this
    email to (in reply to MAIL FROM command)

I suspect the issue has to do with proton announcing itself as, which has not been hosted on for a while. This configuration is a holdover from when proton's hostname was the bare



Updated by lukeshu about 6 years ago

I've configured it to announce itself using the system hostname (remove the myhostname= setting), and also removed the myorigin setting, as it was set to the default anyway.

I've also changed from a CNAME to an A record (having it as an A record is an error, in fact it is one of the common errors listed in RFC 1912 ).

This means that there are 3 A records:


I have also ensured that there are TXT-record and SPF-record SPF settings for those 3 domains, and only those 3 domains.

With the exception SRV records and CNAME records, these are the only domains that there should be any records for.

I've also emailed Jon asking him to change the reverse-DNS for proton's IP from to


Updated by lukeshu about 6 years ago

I've now applied the same treatment to the The only non-CNAME, non-SRV records there must be on the bare domain (which was also erroneously CNAMEed).

I've also altered existing HINFO records, and added any missing HINFO records. Previously, they were all "Parabola" / "Project". The format of HINFO is "processor" / "operating system", where "processor" is nominally one of the IANA registered machine names ( ), and the "operating system" is nominally one of the IANA registered OS names ( ). Of course, neither registry has been up dated in a while (2001 and 2010, respectively), and saying the OS is "LINUX" just won't do. So I've gone with machine=x86_64 and os=Parabola_GNU/Linux-libre.

I've also set the TTL back up to 24h, after having shortened it during the long outage.

Also available in: Atom PDF