Milter-greylist

manu@... wrote:
> And by the way, I'm a bit tired of implementing stuff for which
> there is no feedback at all (DKIM, p0f)...

Emmanuel:  We are all grateful for milter-greylist.  Your devotion to
adding new features is a great boon for everybody.  I often proudly
talk about my use of milter-greylist.  It is extremely robust, easy to
use, and the features are extremely extensive, especially with the
recent additions of DKIM and p0f.  Thank you.


I see greylisting as having two main merits:  evading zombies (and
other non-SMTP compliant servers) and delaying possible spam.
Delaying mail lets others receive and report it first, so it hits the
RBLs and similar online databases before I check the content.  SPF and
DKIM do not appear to do either of those two things, but p0f does - it
allows me to specifically delay Windows servers since they're more
likely to be zombies.  SPF and DKIM come into my spam-fighting picture
later, when SpamAssassin is unleashed on the message.

I plan to give p0f a whirl very soon.  I was the original requester,
though I think it was somebody else's second round of prodding that
got it implemented.


Getting back on topic to P2PWL:

>> To milter-greylist developers (manu):  does this look worthwhile
>> to implement? given the "peer" option, this is already mostly
>> written.
> 
> What is it?

p2pwl is an auto-whitelist sharing mechanism for greylisting hosts.
It would allow users to share the servers who pass, thus creating a
web-of-trust concept to better facilitate a more unilateral system for
dealing with grey lists.

Like p0f (actually, more than p0f), p2pwl is a tool for greylisting,
devised to help greylisting servers be more effective.  Currently, it
appears that only postgrey works with p2pwl.

I think p2pwl is a good idea, expanding on an idea already implemented
for milter-greylist.  This latter fact should make implementation
pretty easy (says a non-developer), especially since the more
complicated stages of p2pwl's development (which offer features not
yet exhibited by milter-greylist) have not yet been finalized.

p2pwl's full power is still in draft form, so perhaps my request is a
bit early.  Stage I, the only completed stage, has manual peer
selection, which is identical to milter-greylist's peer configuration
option.  (See http://oc-co.org/p2pwl/#stages for detail.)

By implementing p2pwl's sharing mechanism instead of a protocol
exclusive to milter-greylist, you open the software to the ability to
share with similar products, perhaps gaining visibility for the
project and maybe even some converts, plus you make the ability to add
those later stages of p2pwl's master plan more possible.


Let's back up and re-visit my original email -- I wanted to know if
anybody had already used p2pwl, and/or if it seems like a good idea.

-Adam