POPFile 0.21.0 has been released, and with very little trouble I have it up and running in Mac OS 10.3.

The big change in this version is no more BerklyDB back end, POPFile now using a SQL database, and the easiest way to do this in OS X is to use SQLite. Thankfully Michael Tsai (SpamSieve author) has a built version of SQLite for Mac OS X. This means you do not have to compile and install it. The schema is available to use something like MySQL, but why bother?

Next up you have some perl modules to install:

• DBI
• DBD::SQLite
• Digest::Base
• Digest::MD5
• HTML::Tagset

These you will need to download, compile, and install. DBI needs to be installed before you can install the DBD::SQLite module.

More on installing SQLite and these associated perl modules below.

There are a couple really cool new features to POPFile, and in my opinion, worth the work to use them. They are:

• The global options for Subject Modification, X-Text-Classification insertion and X-POPFile-Link insertion have been removed and replaced with individual options on a per bucket basis to give greater choice in configuring POPFile.
• The 'unclassified' bucket is now visible in the UI so that you can see how many messages were unclassified, and configure header modification. This also means that unclassified messages are counted in the accuracy statistics; previously they were not counted which could have skewed the accuracy statistics if there were unclassified messages. — This was something I had asked for, and I assume others did as well.
• The history "page" bar has been simplified so that it uses a fixed amount of screen space, while making navigation easy. Filters and searches on the history are now persistent, for example you can click on the Buckets page and return to the History page without losing your filter or search settings.
• The Buckets page has been modified to only show the 'distinct word' count per bucket and to show the total number of distinct words in the database. Previously we showed two counts with confusing titles: now we show the true number of words in the database, not the "word counts" (which was the number of times each word occurred).
• We've recently seen spam start to use CSS to obscure messages and fool filters like POPFile; in response, this version of POPFile does analysis of CSS in HTML encoded messages. POPFile now correctly uses the SpamAssassin headers to make POPFile more efficient when used in conjunction with SpamAssassin. We now also look at TLDs (Top Level Domains) and store them as pseudowords (most useful for TLDs like .biz). — My favorite new feature.