Modularity

[David Relson]

At 09:34 AM 1/13/03, Adriano Nagelschmidt Rodrigues wrote:

>Matthias Andree writes:
> > Not quite. While -f and -r (Fisher-Robinson) and (Robinson) share their
> > registration, -g (Graham) uses a different registration procedure, and
> > the filter algorithm is queried for parameters.
>
>Can they be made common (eg a superset of the information is stored in the
>dbs)?

The difference in registration is very small.  As a message is read, the 
words are collected in a hash table and duplications are counted.  For 
graham, the count is capped at 4 and for robinson and robinson-fisher the 
count is capped at 1.  So wordlists generated with '-r' or '-f' will 
generally have lower counts than wordlists generated with '-g'.  When 
omputing a word's spamicity score, what's important is ratios.  For the 
goodlist and the spamlist each word gets a value based on its count vs the 
message count (for that list).  The word's spamicity score is a ratio of 
the goodlist and spamlist values.  Given all the ratios involved, the 
effects of the caps (1 and 4) is pretty small.  It's not worth worrying about.

>I may be asking naive questions, bogofilter is a blackbox to me.
>
> > OTOH, these programs share a lot of common code (like the tokenizer,
> > data base, and recently, MIME stuff), so that we'd exchange like 10
> > percent of these programs with the other 90% being linked in unchanged.
>
>What about a libbogo.{so,a}?

Part of the build process _does_ create libbogofilter.a which is then used 
in building bogofilter, bogoutil, and bogolexer. It helps simplify the 
Makefile.  We don't distribute the library because there's no need to do so.

> > I wonder if that's worth the effort. It might make the man pages more
> > concise though ;-)
>
>Yes, for example, when I decided to add the '-f' switch last night, it wasn't
>clear to me if I would have compatibility problems with the existent ham/spam
>databases and if I needed to use it with [nNsS].

We're always interested in better documentation.  If you care to rewrite 
the sections that gave you trouble, send us patches.  We'll review them and 
use what's useful.

Ciao,

David