Re: Federated, decentralised communication on the internet (was: domain names, was: hostname)
On Wed, 21 Mar 2018 16:21:44 -0400
Greg Wooledge <email@example.com> wrote:
> On Wed, Mar 21, 2018 at 02:53:47PM -0500, David Wright wrote:
> > HELO dotlessdomainname
> > HELO dotcontaining.home
> > I want someone to explain to me why having a dot is better then not
> > having a dot in deciding whether a submitter is genuine. And
> > without the politics.
> My understanding: the SMTP receiver will use whatever heuristics it
> finds appropriate to avoid receiving spam.
Indeed. Exim4 is fairly easy to configure either way.
> One heuristic that is commonly used is to reject all messages where
> the HELO doesn't even syntactically qualify as a valid FQDN -- in
> other words, has no dot in it.
> Another heuristic that is commonly used is to perform a DNS query on
> the HELO string, and reject it if it's not a valid FQDN based on DNS.
> The first heuristic is much less expensive to perform, as it does not
> involve sending a DNS query and waiting for the response. The test
> is simply a syntactic scan of the input string that it already has.
I'm willing to do DNS lookups. Both the HELO and the PTR record of the
sending IP address need to be resolvable in public DNS for my mail
server to accept mail, as does the domain of the 'envelope' sender. The
latter weeds out a fair bit of spam.
> For the person who is trying to send legitimate outgoing mail,
> obviously you don't want your messages to be rejected as spam. So it
> behooves you to make sure your message complies with not only the
> applicable standards (SMTP = RFC 2821, etc.), but also with the known
> practices of potential receivers. Which means, among other things,
> having a HELO string that won't cause your message to be dropped as
> probable spam.
I haven't done it recently, but I often used to use a telnet connection
to talk manually to many SMTP servers, most of them required a
resolvable HELO, but they didn't care if it had any connection with the
sender. To minimise typing, I used to use a six-character domain name
(there are several) to which I had no rights whatever.
I've seen one of my clients' Exchange servers reject a connection from
BT, the UK telecoms provider, because one of its Exchange servers was
misconfigured to send its own private domain name, with a .local TLD,
as its HELO.