Autor: Peter Bowyer Data: A: Exim, Users Assumpte: Re: [exim] Greylisting algorithms after end of DATA
On 14/01/07, Ian Eiloart <iane@???> wrote: >
>
> --On 13 January 2007 16:20:17 +0000 Peter Bowyer <peter@???> wrote:
>
> >
> > But for most people, running SA is the most expensive test they do,
>
> So, the question is are we doing greylisting to save IT resources, or to
> stop spam from reaching humans (don't forget, they're a resource too!). If
> you don't have the resource to scan all your spam, then you might prefer to
> greylist before scanning.
An unscientific gut-feel is that this is often the case.
> Alternatively, you might be more concerned about
> losing wanted mail as a result of broken or mismatching sender retry
> policies - in that case, you want to avoid greylisting mail that
> spamassassin says is OK.
Of course - one of the unpredictable characteristics of greylisting is
that it depends on a certain class of behaviours from the sending
system, some of which may be compliant but illogical.
What I was trying to work out was the resource load for Magnus's
scheme - if it involves SA scanning a message both before greylisting
and on retry, then it's pretty heavy. But I presume once a message has
passed, the sending MTA is whitelisted...... (modulo hotmail, yahoo,
...)
