From mboxrd@z Thu Jan 1 00:00:00 1970 From: 5srmspw02@sneakemail.com (Guarded Identity) Date: Wed, 26 Mar 2008 00:57:05 -0500 Subject: [sup-talk] New User Questions In-Reply-To: <1199944172-sup-660@south> References: <29393-78803@sneakemail.com> <1199944172-sup-660@south> Message-ID: <13352-37236@sneakemail.com> Hi, I've been using Sup for a few months now, read up a little more on Ruby, and thought I might be in more of a position to respond to this thread. Excerpts from William Morgan's message of Thu Jan 10 00:34:03 -0600 2008: > Excerpts from Guarded Identity's message of Tue Jan 08 19:39:18 -0800 2008: > > Will migration of maildir mail from new/ to cur/ necessitate an execution > > of sup-sync? > > Hm. I don't think so. As long as the mtime and size of the file are > preserved (which I believe they are during a move) you should be ok. I was thinking about the algorithm for calculating the message's unique "source_info" for Maildir mail (mtime appended to filesize). I know altering the backend might really be something you don't want to support, so maybe I'm voicing a moot concern. It's just that sometimes I get annoyed at people sending me mail with large attachments (especially at work where I just can't control this kind of thing too well). In the past, I'd use mutt to purge mail of their large attachments. However, I guess if I do this en mass, I loose my labeling of these messages when I do a sup-sync because the message's source_info changes, right?, but have you considered using a message's "Message-Id" header as part of the source_info? I'm thinking you might have. For the most part, this should be pretty unique, right? I know the Message-Id is sometimes missing, but we can default to the old source_id if things are really wacky (which should be rarely, right?). The only reason I could think you wouldn't use the Message-Id is because maybe it's less efficient to open up a message than look at it's file stats. Or maybe there's another reason? > > I had done this before, actually, but I noticed that Sup indexing was > > slowed down quite a bit (rss2email and mailing lists generate a lot of new > > mail). > > If that's true, then it must be due to IMAP transmission costs. I'm > certainly willing to believe that pulling a message from IMAP is > significantly slower than a disk read. I was thinking about this too. Is there anything we can do to pull down headers first to fill the index (backgrounding the indexing)? Maybe that might make the delay with IMAP less of a nuisance? However, I won't harp on this too much. Because I wasn't comfortable synchronizing (say with Unison) sup Ferret indexes across hosts, I'm ssh'ing into the machine with the index and using "maildir" sources. But this unfortunately causes me a problem with viewing graphical attachments because X forwarding is really slow. So for now I'm saving attachments and scp'ing them over. I'm not sure there's much anyone can do about any of this. > > I would really like to have time-based auto-expiry (excluding starred or > > special-labeled items) for some of my mail (primarily mailing lists and rss > > feeds). With mboxes, I guess I could do some "deleted"-labeling with a Ruby > > script followed by a call to sup-sync-back. > > That sounds like the right approach. > > > 6. What are my scripting options for mail expiry with Maildirs? The search > > to get the message objects is pretty straight forward. How much further > > work is it to delete the message from the index and to get a filename of > > the message to delete from the maildir? > > Trivial, although the filename/IMAP id require minor API changes to expose > those functions. But with a working sup-sync-back that applies to all source > types, you won't have to do anything other than inject :deleted labels > everywhere and simply call sup-sync-back. Okay, So I've been using a helper script that does some things to help me manage my Maildir mail (all other source types are ignored). My primary idea was to have a script to expire mail, but allowing me to save threads labeled "starred" or "save". Eventually I added in a function to manipulate labels. Unlike sup-tweak-labels, this script gathers it's pool of messages from the index using a Ferret query (Chronic allowed) rather than using source URIs. I got around some of the encapsulation inherent in the API by just brute force breaking encapsulation (which Ruby makes kind of ridiculously easy it seems). I've attached the script, and would like to see any feedback if people have some. In particular, I'm eager to know of any concerns people have with the approach or defects people catch. I'm still pretty new to Ruby, so if I did anything flat out stupid with the language, please let me know. By the way, I was thinking that there seems to be a good deal of healthy participation in this list. Does anybody see any value to hanging out in #sup on Freenode? Sometimes it's just nice to have an IRC forum. -Sukant -------------- next part -------------- A non-text attachment was scrubbed... Name: sup-tweak-maildir Type: application/octet-stream Size: 6652 bytes Desc: not available Url : http://rubyforge.org/pipermail/sup-talk/attachments/20080326/cef1c568/attachment.obj