https://bugs.exim.org/show_bug.cgi?id=1885
--- Comment #2 from Thaddeus H. Black <t...@debian.org> --- Jeremy Harris writes: > I'm not quite seeing what you think is a bug in Exim versus what turned > out to be either a bug in your configuration or a problem with your > systems for which you have to enforce workaround handling in your > configuration. Can you be more specific? Thank you for asking. A certain configuration does indeed seem to solve my problem. Maybe your answer will be, "Then just use that configuration!" If so, that's fine with me. But maybe your answer will be, "Exim could have been made more straightforward to configure in this instance." When a third party like Debian has packaged and redistributed Exim, I understand that you cannot always discern whether a user's problem is Exim's bug, is the third party's bug, or is the user's error. This is why I reported my problem to Marc Haber of the Debian Project first. After considering, Marc was not sure, but felt that you might be interested, so I escalate the report to your attention on Marc's advice. Exim handles the mail on both my laptop and my relay host. Postfix is not involved. Before hours of trial and error discovered the secret to tune my configuration, here was the chain of events: [1] Exim on my laptop tried to authenticate itself (STARTTLS/plain) to Exim on my relay server. Usually this worked, but sometimes, sporadically, the authentication either failed or was never tried. [2] Whenever [1] did not work, Exim on my laptop fell back to try unauthenticated relay. [3] Exim on my relay server, of course, is configured to refuse unauthenicated relay. [4] Unfortunately, the result of [3] was a permanent, unrecoverable error 550 on the laptop's end. The 5xx error was gratuitous. It was not necessary. My laptop only needed to retry transmission; but, with a 5xx error, retry was impossible. (One of the solutions I considered was to add 5xx error handling to my copy of Exim's source and to recompile for local use, but this felt wrong, so I did not actually try it.) [5] Incidentally, my relay server (as default-configured) had been trying to relay outward through a nonexistent IPv6 interface. I do not know whether this was relevant to my particular problem, but Rui F. Ribeiro did make some rather interesting remarks in the matter. I have earlier linked his remarks in case you wish to read them. I am inclined to agree with you that this is probably a configuration issue. Indeed, as mentioned in point [4], while investigating the problem, I unpacked your source and studied the 4xx error code (which is clear and easy to read, by the way); but decided in the end that modifying the source probably was not the best way to handle this. Still, you know so much more about the code than I do. It seemed wise to give you the information and let you decide. If this is indeed a configuration issue, though, then the question is whether the problem is best ascribed to user misunderstanding, to infelicity in Debian's default configuration, or to infelicity in your own, original default configuration. Consider: my use case is pretty simple. Its sole nonstandard feature that I know is the lack of an IPv6 interface. It just seems to me that my use case should have been easier to configure. A 10-year Debian Developer but not an experienced mail administrator, I am reasonably agile with configurations and fairly perseverant in efforts to figure things out on my own. I can read and understand RFCs 2821 and 2822, and so on. Eventually, I did figure this problem out on my own, or rather with some kind help from Mr. Ribeiro, as you see. Of course, this could just be a foolish misunderstanding on my part. Have I just invented a hard solution while overlooking an easy one? (I do not have access to my actual configuration at this instant, but if you ask for it, I will append it within the next several days.) -- You are receiving this mail because: You are on the CC list for the bug. -- ## List details at https://lists.exim.org/mailman/listinfo/exim-dev Exim details at http://www.exim.org/ ##
