Changes have been made in CVS for this ticket:

http://cvs.horde.org/diff.php/horde/test.php?rt=horde&r1=1.122.6.29&r2=1.122.6.30&ty=u

Changes have been made in CVS for this ticket:

http://cvs.horde.org/diff.php/horde/test.php?rt=horde&r1=1.179&r2=1.180&ty=u

Makes sense.

Should we add a test for Net_DNS to test.php?

Tweaked and committed, thanks.

Changes have been made in CVS for this ticket:

http://cvs.horde.org/diff.php/framework/Auth/Auth.php?rt=horde&r1=1.142.10.34&r2=1.142.10.35&ty=u
http://cvs.horde.org/diff.php/framework/MIME/MIME/Attic/Headers.php?rt=horde&r1=1.29.10.28&r2=1.29.10.29&ty=u
http://cvs.horde.org/diff.php/framework/NLS/NLS.php?rt=horde&r1=1.82.4.22&r2=1.82.4.23&ty=u
http://cvs.horde.org/diff.php/horde/admin/sessions.php?rt=horde&r1=1.2.2.7&r2=1.2.2.8&ty=u
http://cvs.horde.org/diff.php/horde/config/conf.xml?rt=horde&r1=1.74.2.83&r2=1.74.2.84&ty=u
http://cvs.horde.org/diff.php/horde/docs/CHANGES?rt=horde&r1=1.515.2.521&r2=1.515.2.522&ty=u
http://cvs.horde.org/diff.php/horde/docs/INSTALL?rt=horde&r1=1.90.6.22&r2=1.90.6.23&ty=u

Changes have been made in CVS for this ticket:

http://cvs.horde.org/diff.php/framework/Auth/Auth.php?rt=horde&r1=1.202&r2=1.203&ty=u
http://cvs.horde.org/diff.php/framework/NLS/NLS.php?rt=horde&r1=1.128&r2=1.129&ty=u
http://cvs.horde.org/diff.php/horde/admin/sessions.php?rt=horde&r1=1.17&r2=1.18&ty=u
http://cvs.horde.org/diff.php/horde/config/conf.xml?rt=horde&r1=1.250&r2=1.251&ty=u
http://cvs.horde.org/diff.php/horde/docs/CHANGES?rt=horde&r1=1.1221&r2=1.1222&ty=u
http://cvs.horde.org/diff.php/horde/docs/INSTALL?rt=horde&r1=1.135&r2=1.136&ty=u

here is the patch against FRAMEWORK_3 from cvs.

I did look and it seemed mostly cosmetic.  Looking closer, I see that 
some of those places provide more useful data to users.

Of course, did you take a look at places where this patch works? Most 
of those are externally provided hostnames, e.g. when doing country 
lookups etc.

Does having the hostname, instead of just the IP address, have enough 
value to justify the multiple possible delays?

I personally have temporary timeouts for whatever reason here. An 
immediate reload is successful most of the time.

Is there a real advantage to bothering to retry?  Is it more likely to 
work if tried again?  I think we are just trying to avoid waiting for 
a DNS server that will never respond, so a single try, with any 
timeout shorter than 60 seconds would put it in a better situation.

here is an updated patch, with the defaults set to 2 seconds. 
Personally I think that a name server should always have a fast 
internet connection and the connection between the server running 
horde and the name server will normally also be quite fast.

Defaulting to a 1 second timeout and 1 retry only will result in 
failed hostname lookups on slow connections. Wouldn't it be better to 
use more conservative defaults (5 seconds, 3 retries) and have 
administrators lower these values should they feel this provides a 
better performance?

there was a typo in my net_dns.patch please use the net_dns_v2.patch

attached is a patch which makes the use of Net/DNS optionally. If the 
module is not found it falls back to the php gethostbyaddr. The patch 
also makes the timeouts configurable in horde.

The patch obsoletes the previously uploaded patches.



Is there any chance to have it included into the code base?

and finally sessions.php.patch

...Headers.php patch

...NLS.php patch

and here is a patch for Auth.php

...and here without setting the RES_OPTIONS env. vars.



<?php

require_once 'Net/DNS.php';



$now = time();



$resolver = new Net_DNS_Resolver();

$resolver->retry = 1;

$resolver->retrans = 1;

$response = $resolver->query('153.19.206.48');

if ($response) {

     foreach ($response->answer as $rr) {

         $rr->display();

     }

}



print time() - $now . "\n";

?>

That's it, if using an external pear module it ok :-)



The following cod works fine:



<?php

require_once 'Net/DNS.php';



$resolver = new Net_DNS_Resolver();

$now = time();



putenv('RES_OPTIONS=retrans:1 retry:1 timeout:1 attempts:1');

require_once 'Net/DNS.php';



$response = $resolver->query('153.19.206.48');

if ($response) {

     foreach ($response->answer as $rr) {

         $rr->display();

     }

}

print time() - $now . "\n";

?>

What about Net_DNS_Resolver?



http://pear.php.net/manual/en/package.networking.net-dns.net-dns-resolver.query.php



See $retry and $retrans

after quite a time of searching for a solution I came also to the 
conclusion that this is the only possible way in php to solve this.

The following posting should also be noted:

http://us2.php.net/manual/en/function.gethostbyaddr.php#58970

Something like 
http://us2.php.net/manual/en/function.gethostbyaddr.php#46869 is the 
only thing I see. This feels awfully hackish though.

Well I think not much, except that it looks "nicer" in:



- the "last login" line (see ./framework/Auth/Auth.php)

- in the MIME headers of the sent messages (see 
./framework/MIME/MIME/Headers.php)

- in the active sessions view for the admin (see ./admin/sessions.php)



I guess in ./framework/NLS/NLS.php the IP needs to be resolved, so 
"timeoutable" lookup could help most here.

How much do we get from doing the lookup, as opposed to just showing 
the IP address?

There are some strange providers out there who configure there 
nameserver to block DNS-lookups of IPs in their networks so that they 
run into a time out. This is very annoying since the login into 
Horde/IMP takes very long (> 60 sec.) until the gethostbyaddr() calls 
time out. This leads to strange behavior afterwards e.g. IMP can't 
save sent messages in the sent-mail folder or move messages to the 
trash folder (permission denied).



It would by good to have a dns lookup with a short timeout. 
Unfortunately the php gethostbyaddr() function does not have a timeout 
option.



e.g.

<?php

$now = time();

print gethostbyaddr('153.19.206.48') . "\n";

print time() - $now . "\n";

?>

results in:

153.19.206.48

70