In an earlier note, Bob Proulx said "The best solution for you is the one you understand the best. That is the one you can manage the easiest.” For me, for historical reasons, that has been Postfix.
For several years, I ran a full-fledged Postfix server on a Macintosh running at home. Static IP on DSL. Worked great. About four years ago, the cost to keep the DSL at a decent speed was getting too high so I switched to cable with a dynamic IP and outsourced the mail and web hosting of my domain. But I had processes running on the computer at home that needed to send mail. Easiest thing was to just leave Postfix running and as the cable company does not allow outgoing to port 25, have Postfix relay to my new mail provider using relayhost to the submission port. Other than adding relayhost and a password file referenced by smtp_sasl_password_maps, the only other change I needed to make to Postfix was to add Cyrus SASL (I has been using dovecot for smtpd but only Cyrus is supported for smtp (client)). Even though my computer at home is now on dynamic IP, it has a host name in my domain. The IP address has only changed once in those four years and one of those processes lets me know if it changes so I can quickly update DNS. Most of the processes on my computer send via the Postfix sendmail command although there is one that sends via SMTP so having a local STMP daemon is important (it looks like MSMTP that Peter recommended only works as sendmail command replacement). I’ve only had one issue which is one of those processes at home tries to send me a text message via T-Mobile’s email to text gateway (send email to phonenum...@tmomail.net). At some point in the last year, they started detecting that the mail was being double-relayed (home to mail ISP and the mail ISP to them) and rejecting it. My workaround is to have that process send directly to my mail ISP via CURL but that’s error-prone as a network outage will cause it to fail rather than being held for retry (but since this process retrieves mail from the mail ISP via fetchmail, analyzes it for some keywords, and immediately send the mail via CURL, the outage would have to happen in that fraction of a second between fetch and send). But I just tried it right now via the Sendmail command and it worked so maybe T-Mobile realized that this was rejecting too much legitimate messages. -- Larry Stone lston...@stonejongleux.com
