Re: [docbook-apps] ezmlm has passed on a virus to me

[Noah Slater <nslater@tumbolia.org>]

On Mon, Dec 01, 2008 at 11:27:29AM -0800, Brooks Moses wrote:
> Ron Catterall wrote, at 12/1/2008 10:28 AM:
> > Surely it is not too difficult to scan a failure message to look for
> > strings containing things like parts of:
> >  >>>>
> > SMTP error from remote mailer after end of data:
> >    host mailstore1.secureserver.net [64.202.166.11]:
> >    554 The message was rejected because it contains prohibited virus or
> > spam content
> > <<<<
> > and decide it's not a good idea to include this material in another
> > generated email.
>
> Surely not?  That's the same thing generations of AI researchers have said
> about the problem, too.  :) SMTP bounce messages are notoriously hard to
> parse, and non-standardized; you're either asking for a solution that misses a
> substantial fraction of the messages of that sort (and is thus essentially
> worthless) by only handling a finite list of special cases, or else you're
> asking for an implementation of strong AI capable of parsing natural language.

As someone who has, unfortunately, written mailing list software I can attest
that this is such an impossibly hard task, I wasted many months trying to
achieve even a moderate success rate before abandoning the feature all
together. The only thing of value to the SMTP agent is the error code which has
precisely defined semantics as per the RFC. Even this is troublesome to extract.

-- 
Noah Slater, http://tumbolia.org/nslater