W B Hacker wrote:
> Dean Brooks wrote:
>> On Sun, May 10, 2009 at 02:37:20AM +0800, W B Hacker wrote:
>>> Brian Blood wrote:
>>> I get the impression those participating in the thread are not aware
>>> that the Exim reverse_host_lookup has sufficiently thorough logic to
>>> discriminate between 'good enough' and 'NFW' returns.
>> Neither reverse_host_lookup or the $host_lookup_failed variables are
>> acceptable for testing explicitly for the existance of PTR records.
>> Both will fail if the forward lookup fails.
>>
>> Also, those mechanisms do not give enough granularity over DNS
>> timeouts on *just* the PTR lookup itself. The $host_lookup_deferred
>> will be set if either the PTR or the forward lookup times out, which
>> compounds the problem further.
>>
>> I am unaware of any other way to do this without using dnsdb.
>>
>> --
>> Dean Brooks
>> dean@???
>>
>
> I believe your last statement.
>
> That doesn't flag it immutable.
>
>
> What do you suppose this does?
>
>
> # C_1 PTR RR Test
> #
> warn
> condition = ${if eq{$sender_host_name}{}{yes}{no}}
> log_message = C_1 Null return for PTR RR search. KMELHEA
>
> The real 'experts' here can probably do that in 1/10 the byte-count...
>
> Bill
>
.. or at least fix it 'coz it was the wrong one..
;-)
THIS is what works:
# CONNECT_3A BFBI test for outright absence of a PTR RR
#
deny
logwrite = C3A Checking for basic presence of PTR RR
condition = ${if eq{$interface_port}{25}}
!hosts = : +relay_from_hosts
condition = ${if and{{def:sender_host_address} \
{!def:sender_host_name}}{yes}{no}}
log_message = C3A No PTR RR found for $sender_host_address
NB: key conditional shamelessly stolen from a post by Marc Sherman 3+
years ago:
http://www.mail-archive.com/exim-users@exim.org/msg08258.html
Exemptions for our non-MTA boxen's cron reports and port 587 submission
inferred.
TESTED today to NOT trigger on 'generic' PTR RR, NOR on 'stale'
(once-valid, but now 'orphaned') PTR RR's.
The 'full' reverse_host_lookup test series has actually been invoked to
derive the above, (see ~/src/hosts.c)..
.. but full results are not tested until a subsequent acl clause to
determine if the PTR RR found is of any actual *use* in an smtp environment.
A more all-around useful test, of course.
That said, no 'extra' code is needed, there is a modest saving in
session time and log verbosity by bailing-out earlier.
Risk of falsing due to DNS response delays has yet To Be Determined
(here, anyway).
Bill
