33 ms·

I have 23+ years worth of e-mail. Started as Pine back in 1995, moved to Postmaster (on BeOS) at some point Eudora, Evolution (on Linux), Thunderbird, GyazMail,

I have 23+ years worth of e-mail. Started as Pine back in 1995, moved to Postmaster (on BeOS) at some point Eudora, Evolution (on Linux), Thunderbird, GyazMail, Mail.app and now considering moving to Mutt and pop-ing down mail locally again. Currently takes up about 46gb. I did at one point start to write my own e-mail client and I stored all of my mail in SQLite, using SQL to organize, search, tag, etc. About 4 years ago I had an old employer e-mail me out of the blue asking if I remembered some code that I wrote. I hadn't worked for them in about 15 years. I was able to send them all the e-mails I had from the project that spanned 6 months.

dfee 9y ago
Writing an email client and storing it I SQLite seems easier today than ever before. The challenge seems to more on the side of idiosyncrasies of getting other servers to give yours authority. Very interesting!
- jason_slack 9y ago
  I actually decided to take a look at the project this morning. It still builds on OS X 10.13! I used QT for the UI and all underlying code is portable as well on POSIX systems.
stevekemp 9y ago
> I did at one point start to write my own e-mail client and I stored all of my mail in SQLite, using SQL to organize, search, tag, etc. That sounds like a fascinating idea. I wrote my own email client, but it uses Maildirs for storage (with optional IMAP support), and one of the biggest issues is that reading every file in a large directory is slow. (Well if I'm honest the issue is reading and parsing every file from a given maildir hierarchy.) Caching headers helps some, but if it were all in a database I'm sure it would significantly simpler. "notmuch" went down that route, and it seems like it pays off in terms of speed. The issue really is importing the mail, and making sure your database doesn't get out of sync with the files on-disk. If you were really brave of course you'd remove the maildir-contents post-import. But I think I'd be too scared to do that!
- jason_slack 9y ago
  > If you were really brave of course you'd remove the maildir-contents post-import. But I think I'd be too scared to do that! Oh, I did! I figured why not? It would just mean that I needed to use my client for history if I ever stopped developing it and went back to some other solution. Of course I used a copy of my e-mail while developing and testing, etc.

dfee 9y ago

Writing an email client and storing it I SQLite seems easier today than ever before. The challenge seems to more on the side of idiosyncrasies of getting other servers to give yours authority. Very interesting!

jason_slack 9y ago
I actually decided to take a look at the project this morning. It still builds on OS X 10.13! I used QT for the UI and all underlying code is portable as well on POSIX systems.

stevekemp 9y ago

> I did at one point start to write my own e-mail client and I stored all of my mail in SQLite, using SQL to organize, search, tag, etc. That sounds like a fascinating idea. I wrote my own email client, but it uses Maildirs for storage (with optional IMAP support), and one of the biggest issues is that reading every file in a large directory is slow. (Well if I'm honest the issue is reading and parsing every file from a given maildir hierarchy.) Caching headers helps some, but if it were all in a database I'm sure it would significantly simpler. "notmuch" went down that route, and it seems like it pays off in terms of speed. The issue really is importing the mail, and making sure your database doesn't get out of sync with the files on-disk. If you were really brave of course you'd remove the maildir-contents post-import. But I think I'd be too scared to do that!

jason_slack 9y ago
> If you were really brave of course you'd remove the maildir-contents post-import. But I think I'd be too scared to do that! Oh, I did! I figured why not? It would just mean that I needed to use my client for history if I ever stopped developing it and went back to some other solution. Of course I used a copy of my e-mail while developing and testing, etc.