> On Oct 18, 2018, at 11:32 AM, Bill Cole
> <[email protected]> wrote:
>
> This points out the basic difficulty in doing MTA surveys. It's impossible to
> know what MTA you've connected to in many cases, it is often impossible to
> know if it's the same instance you've already connected to on a different IP,
> and if (as it seems for the above survey) you don't de-duplicate targets by
> IP but only by name, you may talk to the same MTA on the same IP many times
> over by chasing a bunch of different MX names. Beyond that, you also have a
> reverse problem of very large mail hosters masking scores or hundreds or
> maybe thousands of MTA instances behind some combination of round-robin DNS,
> load balancers, and anycasting making them look like a handful of machines.
> And finally there is the Zen question of MTAs: if a Plesk-managed Qmail
> instance that hasn't been able to deliver mail for a decade is the MX for a
> domain with no addresses that anyone cares about, is it really an MTA?
To determine MTA market share, one might see whether Gmail, or similar are
willing
to take a random large corpus of messages they received, look at the topmost
external
"Received" header, extract the MTA name from the "with" clause and publish the
results.
This would produce a mail volume weighted metric. They could also de-duplicate
by the
name in the "by" clause, producing something like a count by MTA host.
Not sure whether any of the large providers would be willing to do that.
--
Viktor.
