[Spambayes] Whitelisting for spam reports

Tony Meyer tameyer at ihug.co.nz
Sun Oct 16 00:22:47 CEST 2005


> Yes, but this way i have always to look on the review page if there  
> is a
> spamcop reply either in the category ham, in spam or in unsure and  
> have
> to select "discard". I 'd like to have a possibility to "auto-discard"
> these emails in Spambayes. I could also call this "ignore list".
> It is a function that would make reviewing easier for the user.

This isn't whitelisting, though, it's something else.  Even if  
SpamBayes included whitelisting, there's no reason to think that  
whitelisted messages would be excluded from the sb_server review page.

Assuming that you're training only on any mistakes & unsures, then  
both ham and spam should be set to default to 'discard' anyway.  Is  
it really that much of a problem to have these in the list when they  
don't cause any extra work?

> To comment the FAQ:
> - as you can see, Spambayes needs in some cases a whitelist.

There isn't agreement about this.

> - ignoring user requests is never a good idea

Open-source is a scratch-your-own-itch world.  It's not like we have  
a great deal of time to work on SpamBayes, or that we're getting any  
financial reward out of doing so.  (Having more non-contributing  
users is in some ways a net negative, because it just adds to the  
support load).

> - and, frankly, it isn't hard to code a simple whitelist for a  
> developer
> that manages to deal with a program like Spambayes

So write one and submit a patch.  However, you're wrong.  Have you  
read the comments from Mark that are linked to from that FAQ entry?   
They discuss the difficulty of adding whitelisting (he's talking  
about adding to the Outlook plug-in, but it mostly all applies to  
sb_server as well).

> The summary of that FAQ answer is simply "No, we don't want a  
> whitelist
> and we are not going to code one." I'd rather accept this answer  
> instead
> of the lengthy excuses.

A better summary would be "Whitelisting is a flawed technique and  
would generally make results worse; since open-source relies on user  
contributions, and the developers have no interest in adding  
whitelisting, this is unlikely to be done.  If a user wished to  
supply a patch, it would probably be accepted; however, adding this  
feature is fairly complicated for these reasons..."

=Tony.Meyer

-- 
Please always include the list (spambayes at python.org) in your replies
(reply-all), and please don't send me personal mail about SpamBayes.
http://www.massey.ac.nz/~tameyer/writing/reply_all.html explains this.




More information about the SpamBayes mailing list