From: Matthias A. <mat...@gm...> - 2024-10-30 19:35:57
|
Am 30.10.24 um 17:06 schrieb Ken Moffat via Fetchmail-users: > Hi, Hi Ken, thanks for reaching out. > > I see that fetchmail no longer supports using an MDA as SMTP > fallback. Yes. That only means there is no fallback from SMTP to MDA if the former fails. The plan is to always behave consistently and not try to do A today and in adverse situations we try B and you have to start looking where/how your mail ended up, because B wasn't equivalent to A (and fetchmail cannot prove or disprove that). I don't like such surprises. So I threw the fallout behavior out without removing MDA or SMTP. Assume you use Postfix. What if you had procmail as a fallback but Postfix uses its own local(8) without procmail? Then procmail would behave differently from Postfix's local... or some messages would be logged by Postfix, others would not. You configure an mda or use --mda on the command line, fetchmail uses MDA, you don't use mda or --mda then fetchmail uses SMTP (default) or LMTP or whatever else you configured. > Sending mail to lists, and getting the incoming, for a home user has > become increasingly complex and I try not to change my setup. As > best I can understand (several years since I had to change it apart > from a password), I'm doing the following: (one human user on my > systems). > > 1. postfix on home server, running as servername.mydomain. > > 2. outgoing: postfix connects to my ISP's server using Cyrus-SASL. > > 3. incoming: a cron job to run fetchmail authenticating to my ISP > and retrieving POP3. > > I have always specified --enable-fallback=procmail, but I see that > my fetchmailrc has a comented-out invocation of procmail with a > comment that procmail is not needed for that if the mail server is > correctly setup. > > Am I worrying unnecessarily ? Maybe. As long as you notice in some - for you, anyways - reasonable timeframe that your Postfix is down, this should be good. I found that on reasonable distros (I have used Fedora Linux and FreeBSD in the past years and have been quite content) there aren't Postfix surprises for me. ...but fetchmail should notice that it has not succeeded in retrieving those messages and will therefore not delete them and retry them later (except maybe if it uses LAST on your POP3 server, in that case you could specify --uidl or uidl). > I see that procmail was originally specified so that if the mail > server was not responding, mail would be processed by procmail. I > assume (if I do not need to change anything) that if my local server > does not respond then the mail will just be queued ? If your default is to talk to SMTP, it's re-fetch the same message later, do not delete it. > > En passant (testing configure on a desktop without kerberos): > CFLAGS: -g -O2 -I/usr/kerberos/include > looks a bit odd : /usr/kerberos ? Yeah, we've had such stuff needed in various circumstances, ./configure adds it, either because it's one of the search paths you have for a ./configure --with-kerberos like option, or as a leftover workaround from Red Hat Linux 9 (must have been in the early 2000's) I forgot to remove. At that time, some SSL implementations needed such an option to find krb5.h. This shouldn't be in use today, but also in /usr/kerberos/include there shouldn't be stuff that disturbs us. Too late for 6.5.0 now, should've known during the betas then I could have done something... We'd need to look at your config.log and/or console output, together with configure.ac, to see what's there, but if fetchmail behaves normally and fetchmail -V doesn't emit strange things, you're probably good. HTH Matthias |