[Pan-users] Re: Filtering on news path

[Mark Eggers]

On Thu, 06 Jan 2005 16:16:34 -0700, Duncan wrote:

> If I'm wrong, someone will no-doubt correct me, but I don't believe the
> path header is normally part of the overview.  The limited headers of
> the overview are unfortunately the only part of the message PAN can
> score or filter at this point.

After looking at the code, I agree.  There is a list of keywords Pan
currently recognizes as items to score on, and it will generate an error
message about others.

> He's back to developing now, but in the mean
> time, others had been working on another major feature, the switch to a
> decent database (sqlite library) backend, and after a quick maintenance
> release likely sometime this quarter, integration of that is likely to
> be the next major project.

I think that's a good idea.  One of the issues I have with Pan is memory
consumption when you have a lot of articles in a particular newsgroup. 
After a while the memory utilization gets pretty unpleasant.

> 
> That said, Charles has always said "patches welcome".  If you are
> looking thru the source that implies you have at least some skills in
> the area I don't.  That would be one patch I'd consider applying here,
> before it made CVS!  I could DEFINITELY use it, and so could a number of
> others who've made similar requests.  Therefore, if you feel inspired,
> please hack away.

I'll certainly take some time to look at it although I can't promise
anything.  I'm trying to write a good modular skin for Forrest
(forrest.apache.org), a ton of how-to documents, and a project management
workbench based on Xindice (xml.apache.org/xindice) and hsqldb.

I'm also pretty frantically looking for work, but that's another story
entirely.

> It's actually likely to make it into CVS as well, assuming a well made
> patch that fits well with the current code, given Charles' past
> invitations.  He's generally been fairly helpful on the developer list
> as well, and like I said, he's around again, so if you have any
> questions about implementation or something, ask away over there, and
> see what transpires.

Thanks!  Once I get a better handle on what the filtering / scoring code
does, I'll give it a shot.  The last thing any developer needs is a
person who hasn't done the homework to ask generalized clueless questions.


> Alternatively, I've toyed with the idea of downloading messages, then
> shutting down PAN and running a script to do my required filtering, then
> starting PAN again and getting back to work.  However, I've never
> actually written such a script.

That can get messy.  I'm mostly concerned with spam in the technical
newsgroups.  The only pattern I've found that has been consistent is a
marker in the Path: header element.

I've not looked at other newsreaders in a while.  Like you, I've become
comfortable with Pan and think it's a pretty nice tool.  I use it for both
binary and text newsgroups, and have not had too much difficulty.  A quick
look at KNode reveals no mechanism for filtering on Path: elements. 

Sounds like an opportunity.

/mde/
just my two cents . . .