On 16Oct2020 00:15, Ranjan Maitra <maitra@xxxxxxxxx> wrote: >On Fri, 16 Oct 2020 10:14:48 +1100 Cameron Simpson <cs@xxxxxxxxxx> wrote: >> Mostly the races. >> >> To add, remove or flag in MH probably requires a locking mechanism while >> updating the message number lists (and correspondingly, allocating new >> message numbers). You also don't know an arriving message is complete unless >> it isn't numbered yet; I imagine an MH insert goes >> save-completely-to-tempfile, allocate-number, rename-tempfile-to-number, >> update-number-lists. A lock would need to be held over the last three >> steps. > >This may be a stupid question, but does this problem still happen if I use fetchmail to pull mail and store messages in files via procmail? Well, it is an issue regardless of the tool. procmail may do the initial placement, but your mail reader probably moves messages around, if only from one folder to another, which is the same task (new message in target folder). >In my experience, messages have been given unique numbers in their respective folders. (Of course, my MH is really not MH, but rather sylpheed-mh, since they do not update the .mh_sequences but use .sylpheed_mark. Well, the requirement for MH is of course that they have unique numbers because their filenames are so named. The important thing is that delivery tools cooperate - more than one tool might be trying to deliver to the folder at once, and only one can work on the .mh_sequences or .sylpheed_mark files at a time. So: locking. Maildir lets you do this without locking because of the noncolliding filename approach, and the "is this message file complete?" issue by doing all preparation in the tmp subdir, before renaming the completed file into the new subdir. >> Maildir is race free. Messages get unqiue filenames (composed of various >> sufficently unique values combined), are created in the "tmp" subdir, >> and renamed into the "new" subdir. Read messages are renamed from "new" >> to "cur". No shared number lists, no locks. You only look for messages >> in "new" and "cur". > >Thanks very much for this. I have found a tool that can convert mh to mailbox: > >https://github.com/vuntz/mh2maildir/blob/master/mh2maildir > >It seems to work, but can not handle a second level of subfolders: brings them all out as individual folders at the first level, so Ihave to fix that. Also, I don't like the new folder names, seem too unnecessary for me. (I was expecting to the old MH folder names inside my Maildir.) Also, the mails get stored as something like: 1602799622.116065_21187.hostname:2, not sure if this is the recommended way that files are stored in the Maildir format. I was expecting to have something that I could have control over. Yah. I wrote one when I made this switch: https://hg.sr.ht/~cameron-simpson/css/browse/bin/mh2maildir?rev=tip Hmm, some years ago now, looking at the opening comment. And I'm using procmail for the conversion (!!!), so indeed quite a while ago. This script moves the MH folder sideways and makes an empty Maildir in its place, then delivers every message from the MH folder into the new maildir. These days I'd use mutt for the bulk conversion instead of procmail. You can see an example of that approach in this script: https://hg.sr.ht/~cameron-simpson/css/browse/bin/mboxify?rev=tip >I have to look into this some more. I am not sure if this is the standard way to store Maildir format messages. mutt doesn't care - there's no "standard". There was a recent discussion on mutt-users with another user moving to Maildir, who had nested folders. He's still got nested folders, exactly as before. Just don't name a subfolder like one of the three reserved names: "tmp", "new", "cur". So do it how you like it. >One aspect of MH that I have liked is that I pull mail on two machines (using fetchmail via a POP server) and they are assigned the same filenames (numbers). Then, if I use rsync with delete, I can delete the corresponding message in the remote machine if I have deleted it on my local machine. It has worked like a charm over the past 15 years (I would say). Ah. This is a little trickier with Maildir, because message flags are stored in the filename. If you change the flags on both machines without an rsync in between you might run into trouble. With MH the flags are in the .mh-sequences file IIRC. You've still have an equivalent problem though - I presume you're excluding the .mh_sequences files from the rsync? Is one of your machines considered the "main" machine where you read and maintain email, and the other a backup? Or do you delete at both ends? I use getmail via a POP server to collect my email to my laptop. I use rsync with Maildirs in my backup process to the home server backup volume. >Thanks, I think that I am also coming round to the view that I should stick to mutt. The listed neomutt features are: https://neomutt.org/feature. I do not know what is not also in mutt. But in any case, I probably won't know the difference. (I don't completely understand all the features.) Mainline mutt has the sidebar patch and I think the trash folder. I'm unsure about the rest. I index my mail with notmuch, but outside of mutt. >I want to read my e-mail locally. Basically, run fetchmail or mbsync in the background to get mail and store in local folders at intervals, then have mutt read them, etc. I like storing my e-mail locally so that I can get to it offline. Same! So I run getmail every 30s to poll my mail and pull it locally. I have postfix configured on my laptop, so sending email is just done with sendmail. If I'm online it goes straight out, and if not it queues until I am online. Cheers, Cameron Simpson <cs@xxxxxxxxxx> _______________________________________________ users mailing list -- users@xxxxxxxxxxxxxxxxxxxxxxx To unsubscribe send an email to users-leave@xxxxxxxxxxxxxxxxxxxxxxx Fedora Code of Conduct: https://docs.fedoraproject.org/en-US/project/code-of-conduct/ List Guidelines: https://fedoraproject.org/wiki/Mailing_list_guidelines List Archives: https://lists.fedoraproject.org/archives/list/users@xxxxxxxxxxxxxxxxxxxxxxx
