On 2023-06-22 02:05:40, Sebastian Nielsen via mailop wrote:
> >> They were going to get a 4xx anyway. I changed the message to *help*
> >> SendGrid.
>
> Yes but if you can change the message for SendGrid only, you can
> accept the mail and let it through... Apparently you were able to
> send custom text to just SendGrid. Then you have some rule to be
> able to differentiate SendGrid mail from other mail. Thus you just
> accept it.
I was unclear then. I patched the mail server to change the 4xx text
that we send to everyone, but I changed it only to help SendGrid.
I have no way to know ahead of time who is using SendGrid. Some
senders, like GitHub, have dedicated servers and send us enough mail
that, over time, I have been able to whitelist all(?) of them. This
helps but is not enough:
* I can't whitelist a server until we've lost a message from it. The
way I find out that there's a server to whitelist is that a
customer calls to let us know that he's missing an important piece
of mail from it (e.g. the Hershey Park tickets).
* Continuing the example, GitHub is constantly adding/changing
servers. When they do, we're at risk of losing mail until we can
whitelist the new addresses.
* The non-dedicated SendGrid servers cannot be whitelisted, because
almost all mail from non-dedicated SendGrid servers is spam.
* I think you're assuming that most of our 4xx errors are from the
spam filter or from overzealous action on my part to make a point;
they're not. We update kernels, reload AV signatures, have
databases go down, accidentally crash postfix during OS upgrades,
typo config files, etc. All of those result in 4xx errors and
whitelisting does nothing to help.
The bottom line is that avoiding 4xx *entirely* is impossible, even if
I could reduce them somewhat at great expense. If nothing else, then
in those unavoidable scenarios, you have to agree that SendGrid has a
responsibility not to delete important mail. But they do.
_______________________________________________
mailop mailing list
mailop@mailop.org
https://list.mailop.org/listinfo/mailop
