On Sun, Sep 24, 2023 at 09:49:52PM +0100, Polarian wrote:
> > No, the choice should be random, to give messages a decent chance of
> > getting through under various conditions.
>
> Why would you ever want to use a protocol randomly?
Because gives mail the best chance to be delivered, if necessary after
a few retries.
> You have Happy Eyeballs which attempts to use the best protocol by
> which responds the fastest, this makes sense for performance, but
> randomly? How does random help?
That may be fine for web services, but is not a good choice for mail,
because many destinations set fairly stringent connection concurrency
limits, and establishing unused connections is liable to trigger rate
limits that impede mail delivery.
> Say you have an issue with IPv6, and by your words random gives the
> best chance. So lets say its truly random without bias, that would
> imply a 50% chance of each protocol being used, would it not?
>
> That means only 50% of the time, the delivery works, how does this
> solve the solution?
When all connection attempts fail, SMTP mail is queued and retried.
A different choice on retry will eventually get through.
You're new here, while Wietse and contributors have been fine-tuning
Postfix to be a best-in-class MTA for 25+ years. If a particular
address selection approach was chosen, it was carefully considered, and
is a good fit for SMTP mail.
Postfix does meet your requirement of using both IPv6 and IPv4, each
some of the time, when both are advertised. An MTA administrator may
choose to use only IPv4 or to prefer IPv4, in which case Postfix will
try IPv4 only or first. The default is sensibly address-family neutral.
Belabouring this topic further will not be productive:
https://www.youtube.com/watch?v=kHue-HaXXzg
--
Viktor.
_______________________________________________
Postfix-users mailing list -- postfix-users@postfix.org
To unsubscribe send an email to postfix-users-le...@postfix.org
