Re: Feature ideas

On Tue, Dec 14, 2004 at 11:32:08AM -0600, Bill Carlson wrote:
> I was thinking that some additional functions in qconfirm-return might 
> help out. The idea would be that adding a "bad" mode, where an address 
> that bounces confirmations instead of just having the bounce dropped 
> results in the sender address being added to the bad list. Over time, 
> typical spammer addresses will get added to the bad list, which will help 
> keep the pending list shorter.
> 
> Thoughts? I'm willing to code, just don't want to start without some 
> discussion.

I mostly stopped development on qconfirm, but would consider to
integrate patches.  It's not because I don't use it anymore, it does an
outstanding job to keep most of my mailboxes and mailing lists almost
free from unsolicited mail.  It's also not because the way it handles
mail is flawed, rather the opposite; for me it proves the concept.  It's
because of the implementation, if I would start from scratch, I would do
lots of things differently today.

But that most probably will not happen, to work on C/R-alike mail
handling software these days is an ungrateful job.  E.g. most people
seem not to share my opinion that delayed bounces (don't reject while
SMTP, but accept a mail message, and inject a new bounce message) is a
Good Thing.  They're upset because of the many bounces caused by mails
with forged envelopes they receive, and very vocal when complaining.

There's another solution to their problem than refraining from injecting
new bounce messages, and rejecting all unwanted mail outright in the
SMTP session as they demand, see

 http://lists.debian.org/debian-isp/2004/10/msg00064.html
 http://marc.theaimsgroup.com/?l=qmail&m=109904994228159&w=2

If that would be accepted and get current practice, delayed bounces
suddenly become a very powerful tool, not only because you can
completely control the contents of the notification, and so make them
more meaningful or meaningful at all, but also to use them to detect
envelope forgery automatically without any changes to the current
internet mail infrastructure.  Finally the whole delivery confirmation
process can be automated, qconfirm already does this.


 http://smarden.org/qconfirm/technical.html

documents all that, but maybe not as easily understandable as it should
be.

Regards, Gerrit.