LWN: Comments on "notmuch release 0.26 now available" https://lwn.net/Articles/743845/ This is a special feed containing comments posted to the individual LWN article titled "notmuch release 0.26 now available". en-us Tue, 09 Sep 2025 08:52:45 +0000 Tue, 09 Sep 2025 08:52:45 +0000 https://www.rssboard.org/rss-specification lwn@lwn.net notmuch release 0.26 now available https://lwn.net/Articles/744649/ https://lwn.net/Articles/744649/ dd9jn <div class="FormattedComment"> A couple of years ago we did a prototype for KDE using gpg to protect a random access index. This worked using an EncFS container managed by the new g13 tool which encrypted the symmetric key using a gpg public key pair. KDE went into another direction and the code is probably pretty rotten now. Later I changed g13 to work with dmcrypt and that is how I now manage my encrypted partitions. If you are interested please ask over at gnupg-devel.<br> </div> Wed, 17 Jan 2018 21:27:40 +0000 notmuch release 0.26 now available https://lwn.net/Articles/744566/ https://lwn.net/Articles/744566/ dkg you're welcome! I hope you find it useful :) <p> GnuPG strives mightily to be cross-platform, and even offers an filesystem container encryption tool called <tt>g13</tt>, but it is not ready for public consumption. <p> But even if <tt>g13</tt> were ready for prime time, for something like notmuch, where storage access latency is likely to be a critical concern, choosing a platform-specific tool is likely to be faster and more efficient. <p> Finally, there's a question of storage allocation -- if the user has to have a separate filesystem entirely for their mail index, that means that they need to decide ahead of time how big the index can be so that they can allocate the backing storage -- and unfortunately, this isn't really an answerable question. So I'm leaning in the direction of making it fast on at least one platform that i can test with (ext4/e4crypt on GNU/Linux), exposing a minimalist and sensible UI at the notmuch layer, and if other developers that prefer other platforms (APFS on OS X? btrfs on GNU/Linux? ext4 on Android? dm-crypt on GNU/Linux?) want to build out the same functionality, we can sort out how to make it work there. Wed, 17 Jan 2018 15:02:29 +0000 notmuch release 0.26 now available https://lwn.net/Articles/744560/ https://lwn.net/Articles/744560/ mxmehl <div class="FormattedComment"> Thanks for the long-term work you've put into the new decrypt/reindex commands, I really look forward to test the 0.26 with my many encrypted mails.<br> <p> <font class="QuotedText">&gt; That said, for people who aren't sure how to best secure their index, i'd be interested in working on "notmuch lock" and "notmuch unlock" subcommands, which could protect the entire index (and maybe even the entire mail store?) using local filesystem-based encryption. </font><br> <p> That would be great to mitigate the possible risks of storing the decrypted content in the index! I'm no programmer but couldn't GnuPG be the more platform-independent solution which is always well-integrated in most systems? Most privacy-aware people interested in such a feature will already have a gpg key.<br> </div> Wed, 17 Jan 2018 12:59:27 +0000 notmuch release 0.26 now available https://lwn.net/Articles/743966/ https://lwn.net/Articles/743966/ dkg As donio points out below, notmuch leaves it up to the user to secure their index, and defaults to not indexing cleartext at all. <p> That said, for people who aren't sure how to best secure their index, i'd be interested in working on "notmuch lock" and "notmuch unlock" subcommands, which could protect the entire index (and maybe even the entire mail store?) using local filesystem-based encryption. <p> In my speculative designs, this is likely to initially be a simple wrapper around <tt>e4crypt</tt> for those people using <tt>ext4</tt> as their local filesystem (people with other filesystems can suggest other approaches, i guess). I welcome further discussion about this approach, either here or over on <a href="https://notmuchmail.org/mailman/listinfo/notmuch">the notmuch mailing list</a>. Wed, 10 Jan 2018 21:00:03 +0000 notmuch release 0.26 now available https://lwn.net/Articles/743960/ https://lwn.net/Articles/743960/ dkg Hi there! I'm one of the contributors to this feature in notmuch. <p> Many people use notmuch while syncing their local mail storage with IMAP. if you permanently decrypt your local mail storage, then you're likely to push the cleartext back into IMAP on the next sync. <p> Additionally, if you decrypt a message permanently, then your local copy of the message will lose whatever other information may have been available in the encryption layer itself -- for example, you'll lose knowledge of the cipher used for encryption, metadata about other targeted keys, etc. If the message was both signed and encrypted, and that was done at the same "layer" (i.e., it is not MIME-clearsigned message that has itself been encrypted), most existing OpenPGP e-mail toolkits will drop the signature entirely. <p> Speed-wise, if you use the session-key caching features in notmuch 0.26, rendering encrypted mail that has already been indexed is significantly faster than it used to be. I've seen reports of a 8× speedup when rendering large encrypted threads, since no asymmetric crypto is needed. So it doesn't cost much to keep the encrypted copy around and render it in the clear when needed. <p> Finally, it's just generally good archival practice to leave your incoming messages untouched -- you never know what information or metadata you might need from the message. Wed, 10 Jan 2018 20:54:50 +0000 notmuch release 0.26 now available https://lwn.net/Articles/743945/ https://lwn.net/Articles/743945/ donio <div class="FormattedComment"> It's optional and off by default. The documentation and help message for the config option warns the user to consider security implications:<br> <p> index.try_decrypt<br> [STORED IN DATABASE] When indexing an encrypted e-mail message, if this variable is set to true, notmuch will try to decrypt the message and index the cleartext. Be aware that the index is likely sufficient to reconstruct the cleartext of the message itself, so please ensure that the notmuch message index is adequately protected. DO NOT USE index.try_decrypt=true without considering the security of your index.<br> <p> Default: false.<br> <p> </div> Wed, 10 Jan 2018 19:01:47 +0000 notmuch release 0.26 now available https://lwn.net/Articles/743936/ https://lwn.net/Articles/743936/ hkario <div class="FormattedComment"> because it's an indexer, not a MUA, it can't/shouldn't rewrite the files it is processing<br> <p> and in general I'd say it's easier to keep the files in encrypted form than deal with all the immutability issues that decryption and saving the decrypted version would bring<br> </div> Wed, 10 Jan 2018 18:05:51 +0000 notmuch release 0.26 now available https://lwn.net/Articles/743914/ https://lwn.net/Articles/743914/ welinder <div class="FormattedComment"> That doesn't seem to be the case as I read it.<br> <p> It seems they are merging the inconvenience of encrypted mail with the unprotected<br> nature of plain text mail.<br> <p> If there is some kind of trade-off there, I don' t see it. Why keep things encrypted<br> if the contents is leaked via the index?<br> <p> </div> Wed, 10 Jan 2018 16:36:05 +0000 notmuch release 0.26 now available https://lwn.net/Articles/743901/ https://lwn.net/Articles/743901/ bandrami <div class="FormattedComment"> <font class="QuotedText">&gt; It's now possible to include the cleartext of encrypted e-mails in the notmuch index</font><br> <p> Err... I hope the index itself is encrypted, in that case?<br> </div> Wed, 10 Jan 2018 15:56:04 +0000