Max number of messages per connection
Sam Hayes Merritt, III
Using 8.13 with ClientRate and ClientConn. Working good to limit
number of connections within a minute window and number of active
connections.
Is there a way to limit the number of messages per connection?
It appears the spammers abusing proxies have figured how how to
send a RSET and continue sending messages. It'd be great to
also limit the maximum number of messages a connection can send.
Wondering if that might already exist.
Thanks,
sam
Andy
<Sam Hayes Merritt, III <s...@themerritts.org>> wrote in message
news:10aaas5...@corp.supernews.com...
Sam,
I'm sorry as I do not know the answer to your question. I did want to make
a suggestion though.
If you are getting hit by allot of proxy boxes, I suggest using the
xbl.spamhaus.org RBL.
This RBL is wonderful and blocks allot of these connections.
Bfons
Sam Hayes Merritt, III
> If you are getting hit by allot of proxy boxes, I suggest using the
> xbl.spamhaus.org RBL.
This is more for our own customers that are infected with a proxy.
We can limit outbound SMTP to our mail servers, limit the number of
connections per minute, the number of active connections, the next
step would be number of email messages per connection.
Thanks,
sam
Claus Aßmann
> Is there a way to limit the number of messages per connection?
> It appears the spammers abusing proxies have figured how how to
> send a RSET and continue sending messages. It'd be great to
> also limit the maximum number of messages a connection can send.
> Wondering if that might already exist.
Doesn't look like it. Send us a patch :-)
See sendmail/srvrsmtp.c, add a counter for the MAIL command.
The "main" problem is how to set the limit (per mailer, via
a ruleset, global, ...?)
--
A: Maybe because some people are too annoyed by top-posting.
Q: Why do I not get an answer to my question(s)?
A: Because it messes up the order in which people normally read text.
Q: Why is top-posting such a bad thing?
Malcolm Dew-Jones
: Using 8.13 with ClientRate and ClientConn. Working good to limit
: number of connections within a minute window and number of active
: connections.
: Is there a way to limit the number of messages per connection?
Not sure about the regular settings, but it would be pretty trivial for a
milter to do this.
--
(Paying) telecommute programming projects wanted. Simply reply to this.
Sam Hayes Merritt, III
> Doesn't look like it. Send us a patch :-)
We'll investigate that.
> See sendmail/srvrsmtp.c, add a counter for the MAIL command.
> The "main" problem is how to set the limit (per mailer, via
> a ruleset, global, ...?)
It would probably need to follow the same ClientRate where you
can set a global and also change it for a group of IPs.
sam
Andrzej Adam Filip
> Sam Hayes Merritt, III wrote:
>>Is there a way to limit the number of messages per connection?
>>It appears the spammers abusing proxies have figured how how to
>>send a RSET and continue sending messages. It'd be great to
>>also limit the maximum number of messages a connection can send.
>>Wondering if that might already exist.
>
>
> Doesn't look like it. Send us a patch :-)
>
> See sendmail/srvrsmtp.c, add a counter for the MAIL command.
> The "main" problem is how to set the limit (per mailer, via
> a ruleset, global, ...?)
Do you suggest *fixed* limits ?
I expected you to tempt donors into making "load adaptative" limits :-)
IMHO it does make sense to use different limit during peak hours and
overnight "sleep time".
--
Andrzej [en:Andrew] Adam Filip an...@priv.onet.pl an...@xl.wp.pl
http://anfi.homeunix.net/ http://slashdot.org/~anfi
*Random Epigram* :
An egg is dear on Easter day.
-- Russian Proverb
Claus Aßmann
> Do you suggest *fixed* limits ?
> I expected you to tempt donors into making "load adaptative" limits :-)
> IMHO it does make sense to use different limit during peak hours and
> overnight "sleep time".
How do you determine "peak hours"?
Do you suggest fixed times?
Some things don't need to be complicated.
Jochen Bern
> Andrzej Adam Filip wrote:
>>Do you suggest *fixed* limits ?
>>I expected you to tempt donors into making "load adaptative" limits :-)
>>IMHO it does make sense to use different limit during peak hours and
>>overnight "sleep time".
> How do you determine "peak hours"?
> Do you suggest fixed times?
> Some things don't need to be complicated.
Well, I'll leave that discussion to the two of you ;-), but since I'm
preparing a multipurpose patch anyway, I included the checks and
disconnection routines as a first step - with the actual limits given at
compile time, just like the other MAX* macros in srvrsmtp.c .
Announcement in a separate thread.
Regards,
J. Bern
Andrzej Adam Filip
> Andrzej Adam Filip wrote:
>>Do you suggest *fixed* limits ?
>>I expected you to tempt donors into making "load adaptative" limits :-)
>>IMHO it does make sense to use different limit during peak hours and
>>overnight "sleep time".
>
> How do you determine "peak hours"?
> Do you suggest fixed times?
Fixed time periods (e.g. like UUCP time periods specification) or based on
system load (e.g. number of messages received during last 15m).
> Some things don't need to be complicated.
Keeping things simple is not a bad idea but making some thing a little bit
more complicated makes them more useful.
I have assumed that "true goal" of the inquiry is to find how to avoid
sendmail overload (during peak load). Delaying remote sites delivery attempts
from "off peak time" to "rush hours" is counter productive.
We are not GIs in an army. We do do not have to follow orders "as they are".
We are allowed to optimize/question orders themselves.
--
Andrzej [en:Andrew] Adam Filip an...@priv.onet.pl an...@xl.wp.pl
http://anfi.homeunix.net/ http://slashdot.org/~anfi
*Random Epigram* :
Under capitalism, man exploits man. Under communism, it's just the opposite.
-- J.K. Galbraith
Claus Aßmann
> Fixed time periods (e.g. like UUCP time periods specification) or based on
> system load (e.g. number of messages received during last 15m).
How do you want to measure the latter from within sendmail?
> I have assumed that "true goal" of the inquiry is to find how to avoid
> sendmail overload (during peak load). Delaying remote sites delivery attempts
> from "off peak time" to "rush hours" is counter productive.
The original query was about spammers sending many mails per sessions,
hence it would be more a "unknown or bad client" issue.