Training scripts

[Stroller]

On Jan 27, 2004, at 10:40 pm, Matej Cepl wrote:

> ....
>> I'm a little unclear why you appear to be calling KMail, for
> instance,
>> and your use of the `formail` command suggests to me you're
> doing
>> something cleverer than I.
>
> I am not calling KMail at all, just getting rid of some
> additional email headers put there by bogofilter and KMail.

Ah, I see. Bogofilter should ignore its own headers, IIRC.

>> What I've realised, however, is that if I run my script then
> move a
>> message, say, from my inbox (which is ignored in case it has
> spam in
>> it) to a saved items folder, then subsequent runs of my script
> will not
>> train on that message.
>
> You are sure to know by heart these two pages, aren't you?
> http://cr.yp.to/proto/maildir.html
> http://www.qmail.org/man/man5/maildir.html

I'm familiar with the second, but was not with the first. AFAICT, tho', 
neither the content nor the filename of a read message changes when I 
move it from INBOX to Saved Messages (using Courier-IMAP).

The only way I can see to search on such messages is perhaps by using 
`find` and its -cmin / -cnewer / -ctime options. What does a "file's 
status was last changed" indicate, please..? Would this be valid if the 
file was moved..? The modification time is not changed when it's moved 
to a different directory, which prevents the -mmin from showing true.

>> ...with the script the way it is I can move older messages
>> around & completely rebuild my database from scratch (by
>> removing the old one) in less than 5 minutes.
>
> I am retraining just around 10 messages a day in one run, so the
> speed is not so much issue for me.

Well, I only pointed it out in this instance because to completely 
rebuild the DB is a work-around for this minor glitch of mine. That I 
can rebuild the DB in less than 5 minutes means it's no problem to 
rectify the occasional missed messages by doing so; that wouldn't be 
the case if it took 40 minutes.

Stroller.