Mail learning 2: Thunderbird and Postbox

Subject tags: 

Since writing my post the other day on mail learning, I’ve had a few people contact me to say “how about my product?” So for completeness, here’s a look at two more competitors in the arena, Postbox and Mozilla Thunderbird 3.

I’m finding it hard to separate what I want to say about these two, because they’re essentially the same. Postbox is a commercially-branded version of Thunderbird 2, with a few changes - mainly improvements to the UI. One of the big differences between Thunderbird 2 and Postbox is that Postbox indexes emails into a SQLite full-text-search database, and locates and dumps assets from the emails into database tables, and slaps a nice search UI on top of this database.

This is particularly interesting for me because my previous incarnation of the mail indexing idea was being implemented as a Thunderbird plugin, and doing all this in Thunderbird 2 was a nightmare. You had to go through various contortions to even read any mail, and then do complicated mozStorage things to get it into a database. (I even reimplemented a lot of mail-handling stuff in Javascript because there weren’t easy ways to get simple information out of the email.) And then I gave up because I realised that I wanted a proper search engine - SQLite FTS wasn’t there at the time, and I’m still not convinced about using it instead of something like Lucene - and I didn’t want to mess with componentising CLucene since compiling stuff for Mozilla still breaks my head.

So sticking with Thunderbird 2 is a hard thing for Postbox, andseems like a lot of work for nothing, particularly since the most recent betas of TB3 actually do all this already. Latest betas of TB 3 include an optional component called “gloda” - global database - which is a SQLite full-text-search database of all your email, which locates and dumps assets from the emails into a database table - actually an RDF-like triplestore. And it has a decent plug-in architecture for asset discovery - implementing all my preferred asset recognisers in Gloda would actually be fairly simple. (The Postbox preferences do refer to Calais entity types, but as far as I can see, that preference doesn’t do anything yet.)

As far as mail indexing goes, I’m excited about Gloda. Currently Postbox has the edge in terms of UI, but Gloda’s search UI is not that far behind and its indexing infrastructure is much better and will allow Thunderbird to do more interesting things with email in a cleaner and faster way.

Leaving aside mail learning for a second and looking at usability as a mail client, Postbox has a few interesting features. I really like the ability to see mail conversations even if they span multiple folders; integrating the Firefox Facebook Toolbar to get pictures of correspondents is cute and fun, but some components (like the Twitter and Friendfeed ones) seem to be there more to tick the Web 2.0 box than to be actually useful.

It’s going to be a tight race, but for open-source ease-of-development and for the price tag of zero, if you want a mail-learning desktop email client, Thunderbird+Gloda is going to be the one to watch.