-M inconsistency

[David Relson]

On Mon, 05 Jan 2004 13:32:49 +0100
Boris 'pi' Piwinger <3.14 at logic.univie.ac.at> wrote:

> Hi!
> 
> I think this is the time to fix an old inconsistency about
> the need for -M.
> 
> When training with -s or -n, mboxes are autodetected, so -s
> and -sM are the same.
> 
> When checking a message with -v, -M is needed to get
> meaningful output, -v alone treats it as one message which
> does not make any sense. AFAICS it is a recent fix that now
> only -vvM and -vvvM produce something useful.

Not so.  It used to be that mailbox processing happened only during
registration, i.e. -s, -n, -S, and -N.  During scoring, each input file
was treated as a single message and gave a single score.  -M changed the
scoring of mailboxes.

To make it consistant, -M would be necessary when registering mailboxes.

FWIW, "-vM" produces useful output.

The verbose flag, '-v', is totally separate and controls how much info
is generated during processing.

> Same for -T.
> 
> Bogofilter without options on mbox does not make any sense
> and hence should produce an error I think.

Are you suggesting that -M be required when processing mailboxes?

> So what to do? I don't know how autodetection is done, but
> it works, so it should also work for -v and -T (-t, -TT).
> 
> BTW: -t could also go completely.

'-t' is eye candy.  It provides a terse mode of output whose format can
be controlled.  See "terse_format" in bogofilter.cv.example