From: Alex R. <mys...@gm...> - 2015-07-02 00:40:05
|
Hi, It appears the default installation includes an auto-whitelist entry for google.com and yahoo.com. Is there a particular reason for this that would prevent me from removing it? I'm seeing quite a bit of spam that I believe could otherwise be blocked if the whole domain wasn't whitelisted. Ideas greatly appreciated. Thanks, Alex |
From: Karl O. P. <ko...@me...> - 2015-07-02 05:08:43
|
On Wed, 01 Jul 2015 20:39:57 -0400 Alex Regan <mys...@gm...> wrote: > It appears the default installation includes an auto-whitelist entry > for google.com and yahoo.com. > > Is there a particular reason for this that would prevent me from > removing it? I'm seeing quite a bit of spam that I believe could > otherwise be blocked if the whole domain wasn't whitelisted. The google and yahoo servers are going to obey the rules and retry after an appropriate interval and deliver the spam anyway. You may as well reduce the load and accept it immediately. They may or may not also resend from a different box, doing a round-robin on outbound mail gateways. Mail (probably) eventually gets delivered when all the boxes (IPs) are whitelisted. But it means that legitimate mail can get overly delayed. (Dunno who's doing this any more.) Regards, Karl <ko...@me...> Free Software: "You don't pay back, you pay forward." -- Robert A. Heinlein |
From: Alex <mys...@gm...> - 2015-07-03 01:45:29
|
Hi, On Thu, Jul 2, 2015 at 1:08 AM, Karl O. Pinc <ko...@me...> wrote: > On Wed, 01 Jul 2015 20:39:57 -0400 > Alex Regan <mys...@gm...> wrote: > >> It appears the default installation includes an auto-whitelist entry >> for google.com and yahoo.com. >> >> Is there a particular reason for this that would prevent me from >> removing it? I'm seeing quite a bit of spam that I believe could >> otherwise be blocked if the whole domain wasn't whitelisted. > > The google and yahoo servers are going to obey the rules > and retry after an appropriate interval and deliver > the spam anyway. You may as well reduce the load and > accept it immediately. Ah, right, makes sense. > They may or may not also resend from a different box, > doing a round-robin on outbound mail gateways. > Mail (probably) eventually gets delivered when all the boxes (IPs) > are whitelisted. But it means that legitimate mail > can get overly delayed. (Dunno who's doing this any more.) I'm using db_cluster to avoid this. round-robin issue. This was one of the main reasons I chose sqlgrey over the others like postgrey. Is this not what you're talking about? Thanks, Alex |
From: <da...@ha...> - 2015-07-03 09:46:22
|
Hi Alex.. On 2015-07-03 03:45, Alex wrote: > They may or may not also resend from a different box, > doing a round-robin on outbound mail gateways. > Mail (probably) eventually gets delivered when all the boxes (IPs) > are whitelisted. But it means that legitimate mail > can get overly delayed. (Dunno who's doing this any more.) > I'm using db_cluster to avoid this. round-robin issue. This was one of > the main reasons I chose sqlgrey over the others like postgrey. Is > this not what you're talking about? No db_cluster (or a central db for that matter) solves the problem of 1 foreign IP hitting multiple servers at YOUR end. Not multiple foreign servers, with different IP's, hitting your mail server(s). sqlgrey stores the ip of the sender, either full or as partial, in the db. But if that ip changes, sqlgrey sees it as an entirely new "connection" and has to add that one as well. So now sqlgrey has 2 new connections from the same mail-cluster and both new ip's must resubmit to be whitelisted and be able to send mail through sqlgrey. Hence; they are treated as 2 different mail setups. One always has to remember the purpose of SQLGrey & Greylisting. The purpose is not filtering spam. The purpose is filtering out mail-senders that aren't real mailservers. if it IS a real mailserver but it still sends spam, that is NOT something sqlgrey can do anything about. Thats a job for other spam-filtering mechanisms. So, if we KNOW that eg. google.com are actually real mailservers and that they will alway behave as such, then there is no danger of whitelisting, as you are only whitelisting the fact that they ARE real mailservers; not that they arent sending spam. And as Karl points out, you save time for the clients by whitelisting, as they dont have to do the whole resubmit-dance to determine that they are in fact a real mailserver, when we already know it is. > Thanks, > Alex > Regards - Dan |
From: Bruce B. <bb...@bo...> - 2015-07-03 16:38:07
|
Good day, I'm wondering how many of you have implemented the sqlgrey mod described here... http://www.hyllander.org/sqlgrey_whitelisting It's not very often that I receive a call from a customer informing me that the email that he attempted to send to us resulted in an error message (from his perspective). And the last time was when he was replying to a message that I sent to him! Of course, the "auto-whitelisting of recipient addresses" wouldn't help a bit until someone in my domain writes to a customer but it might add an additional layer of automation. I realize that the problem relates to broken mail servers but it's our customers who are inconvenienced and our business that might suffer. Does anyone have experience with this auto-whitelisting of recipients mod? Thanks, ~Bruce |
From: Lionel B. <lio...@bo...> - 2015-07-03 17:05:39
|
Hi Bruce, On 07/03/15 18:13, Bruce Bodger wrote: > Good day, > > I'm wondering how many of you have implemented the sqlgrey mod described > here... http://www.hyllander.org/sqlgrey_whitelisting It's not very > often that I receive a call from a customer informing me that the email > that he attempted to send to us resulted in an error message (from his > perspective). And the last time was when he was replying to a message > that I sent to him! Of course, the "auto-whitelisting of recipient > addresses" wouldn't help a bit until someone in my domain writes to a > customer but it might add an additional layer of automation. I realize > that the problem relates to broken mail servers but it's our customers > who are inconvenienced and our business that might suffer. It's unlikely your business will suffer: if your customer reports this kind of errors, he probably have them with multiple other destinations (unless his/her admin was temporarily playing with parameters of the mail server) as greylisting and temporary errors are commonplace. If you aren't prepared to explain that everything is fine there is support for whitelisting your customer mail servers (instead of his domain or addresses as implemented in the patch above which would make easy to circumvent greylisting by simply using fake addresses derived from your known customers). Just add entries to /etc/sqlgrey/clients_fqdn_whitelist.local or /etc/sqlgrey/clients_ip_whitelist.local depending on how you want to whitelist (by fqdn or ip). You don't have to restart SQLgrey, these files are automatically reloaded when modified. Best regards, Lionel |
From: <da...@ha...> - 2015-07-03 17:26:27
|
Just a side note. On 2015-07-03 18:13, Bruce Bodger wrote: > often that I receive a call from a customer informing me that the email > that he attempted to send to us resulted in an error message (from his > If you by "error message" mean that the sender receives a mail that looks like a bounce (but isn't), I eliminated most of those at my old job, by using reject_code = 451 which is actually why i added the reject_code option in the first place :).. - Dan |
From: Karl O. P. <ko...@me...> - 2015-07-03 17:33:55
|
On Fri, 3 Jul 2015 11:13:03 -0500 Bruce Bodger <bb...@bo...> wrote: > I'm wondering how many of you have implemented the sqlgrey mod > described here... http://www.hyllander.org/sqlgrey_whitelisting Some thoughts, take them for what you will: If it were me I'd avoid modifying sqlgrey and write a little postfix filter that frobs the database whitelist directly. I'm no advocate of re-inventing the wheel, and honestly didn't read the article in detail, but simpler seems better than more complicated. On a related note, to actually solve the problem you must know exactly what the problem is. It's not clear from your description just why there was an issue in the first place. Karl <ko...@me...> Free Software: "You don't pay back, you pay forward." -- Robert A. Heinlein |
From: Bruce B. <bb...@bo...> - 2015-07-03 22:21:20
|
Thanks for your feedback Karl, Dan, and Lionel.. On 7/3/15 12:33 PM, Karl O. Pinc wrote: > On Fri, 3 Jul 2015 11:13:03 -0500 > Bruce Bodger <bb...@bo...> wrote: > >> I'm wondering how many of you have implemented the sqlgrey mod >> described here... http://www.hyllander.org/sqlgrey_whitelisting > Some thoughts, take them for what you will: > > If it were me I'd avoid modifying sqlgrey and write a little > postfix filter that frobs the database whitelist directly. > I'm no advocate of re-inventing the wheel, and honestly > didn't read the article in detail, but simpler seems better > than more complicated. > I understand your perspectives. I've been running sqlgrey since 2009. I was just wondering if you saw this issue as a big deal and, obviously, you don't. As soon as I hear from a recipient having any problem sending us email I check the mail log and then add their domain to clients_fqdn_whitelist.local if I see that they've truly connected. Again, thanks for the replies. BTW, I've kept the reject_code = 451 default since original installation. P.S. Even though I've been running sqlgrey for 6 years I've never been sure what the whitelist requires. Is troubled.com equivalent to *.troubled.com? My intention is to add all servers and users in the troubled.com domain. I've been adding both. Thanks again, Bruce |
From: Bruce B. <bb...@bo...> - 2015-07-23 13:40:39
|
On Jul 3, 2015, at 5:26 PM, Lionel Bouton <lio...@bo...> wrote: > On 07/04/15 00:21, Bruce Bodger wrote: >> [...] >> >> P.S. Even though I've been running sqlgrey for 6 years I've never been >> sure what the whitelist requires. Is troubled.com equivalent to >> *.troubled.com? My intention is to add all servers and users in the >> troubled.com domain. I've been adding both. > > The whitelisting is about the ips or fqdns of the mail servers > contacting yours and not the addresses of the sender (so they might not > share the same domain). If you are not sure about what they are you > should be able to find them in your postfix logs. Lionel, Does this comment from the distributed clients_fqdn_whitelist still apply? # Note you need the following two lines to allow both # <lots of mtas>.example.com and example.com # *.example.com # example.com Thank you, Bruce |
From: Lionel B. <lio...@bo...> - 2015-07-23 16:58:40
|
Hi, Le 23/07/2015 15:40, Bruce Bodger a écrit : > [...] > Lionel, > > Does this comment from the distributed clients_fqdn_whitelist still apply? > > # Note you need the following two lines to allow both > # <lots of mtas>.example.com and example.com > # *.example.com > # example.com Yes it does. To be honest I wasn't happy about this little quirk when I coded the whitelists support and with experience I would not do it again the same way if I were starting from scratch. IIRC (I didn't look at SQLgrey code for quite some time) this is needlessly linked to internal SQLgrey structures involved in the whitelisting: regexp, globs and identity matching are done separately for performance reasons. I think I should have used a cleaner syntax and made the effort to translate it to the internal structures instead of looking for a direct mapping. As it didn't seem too bad and cleaner solutions would have broken backward compatibility this was left like this (and the comment was added to make sure people reading documentation like you were aware of it). Best regards, Lionel |
From: Lionel B. <lio...@bo...> - 2015-07-03 22:26:31
|
On 07/04/15 00:21, Bruce Bodger wrote: > [...] > > P.S. Even though I've been running sqlgrey for 6 years I've never been > sure what the whitelist requires. Is troubled.com equivalent to > *.troubled.com? My intention is to add all servers and users in the > troubled.com domain. I've been adding both. The whitelisting is about the ips or fqdns of the mail servers contacting yours and not the addresses of the sender (so they might not share the same domain). If you are not sure about what they are you should be able to find them in your postfix logs. Lionel |
From: Bruce B. <bb...@bo...> - 2015-07-03 23:44:32
|
On 7/3/15 5:26 PM, Lionel Bouton wrote: > The whitelisting is about the ips or fqdns of the mail servers > contacting yours and not the addresses of the sender (so they might not > share the same domain). If you are not sure about what they are you > should be able to find them in your postfix logs. > Thank you, Lionel. Understood. Regards, ~Bruce |