Re: soft_bounce=yes in postmulti setup
On 8/9/2012 3:06 PM, Wietse Venema wrote: > Noel Jones: >> On 8/9/2012 12:24 PM, Wietse Venema wrote: >>> Noel Jones: On 8/9/2012 11:27 AM, Wietse Venema wrote: > It might be practical to add a note that in some cases, soft_bounce=yes > is implemented by modifying server responses. Therefore, the responses > that Postfix will log may differ from the responses that Postfix > actually sends or receives. This will be sufficient for the few > people that we write documentation for. Indeed, a brief note would have been sufficient in this case. Note: postfix may log a 5xx response before the transformed 4xx response is sent to the client. >>> >>> No, this applies to Postfix client and server code. >>> >>> Wietse >>> >> >> Well, we can pretend we're sculpting and create by removing. >> >> Note: postfix may log a 5xx response before transforming it to 4xx. > > If we wanted that level of detail, then should also have: > > Note: with soft_bounce=yes, Postfix may log a 4xx response after > receiving 5xx. > > The text that I drafted three hours ago avoids this and is generic: > the response that Postfix logs may differ from the response that > Postfix actually sends or receives. > > Wietse > Yes, that does it nicely. -- Noel Jones
Re: soft_bounce=yes in postmulti setup
Noel Jones: > On 8/9/2012 12:24 PM, Wietse Venema wrote: > > Noel Jones: > >> On 8/9/2012 11:27 AM, Wietse Venema wrote: > >>> It might be practical to add a note that in some cases, soft_bounce=yes > >>> is implemented by modifying server responses. Therefore, the responses > >>> that Postfix will log may differ from the responses that Postfix > >>> actually sends or receives. This will be sufficient for the few > >>> people that we write documentation for. > >> > >> Indeed, a brief note would have been sufficient in this case. > >> > >> Note: postfix may log a 5xx response before the transformed 4xx > >> response is sent to the client. > > > > No, this applies to Postfix client and server code. > > > > Wietse > > > > Well, we can pretend we're sculpting and create by removing. > > Note: postfix may log a 5xx response before transforming it to 4xx. If we wanted that level of detail, then should also have: Note: with soft_bounce=yes, Postfix may log a 4xx response after receiving 5xx. The text that I drafted three hours ago avoids this and is generic: the response that Postfix logs may differ from the response that Postfix actually sends or receives. Wietse
Re: soft_bounce=yes in postmulti setup
On 8/9/2012 12:24 PM, Wietse Venema wrote: > Noel Jones: >> On 8/9/2012 11:27 AM, Wietse Venema wrote: >>> It might be practical to add a note that in some cases, soft_bounce=yes >>> is implemented by modifying server responses. Therefore, the responses >>> that Postfix will log may differ from the responses that Postfix >>> actually sends or receives. This will be sufficient for the few >>> people that we write documentation for. >> >> Indeed, a brief note would have been sufficient in this case. >> >> Note: postfix may log a 5xx response before the transformed 4xx >> response is sent to the client. > > No, this applies to Postfix client and server code. > > Wietse > Well, we can pretend we're sculpting and create by removing. Note: postfix may log a 5xx response before transforming it to 4xx. -- Noel Jones
Re: soft_bounce=yes in postmulti setup
Noel Jones: > On 8/9/2012 11:27 AM, Wietse Venema wrote: > > It might be practical to add a note that in some cases, soft_bounce=yes > > is implemented by modifying server responses. Therefore, the responses > > that Postfix will log may differ from the responses that Postfix > > actually sends or receives. This will be sufficient for the few > > people that we write documentation for. > > Indeed, a brief note would have been sufficient in this case. > > Note: postfix may log a 5xx response before the transformed 4xx > response is sent to the client. No, this applies to Postfix client and server code. Wietse
Re: soft_bounce=yes in postmulti setup
On Thu, Aug 09, 2012 at 11:57:00AM -0500, Noel Jones wrote: > On 8/9/2012 11:27 AM, Wietse Venema wrote: > > It might be practical to add a note that in some cases, soft_bounce=yes > > is implemented by modifying server responses. Therefore, the responses > > that Postfix will log may differ from the responses that Postfix > > actually sends or receives. This will be sufficient for the few > > people that we write documentation for. > > Indeed, a brief note would have been sufficient in this case. > > Note: postfix may log a 5xx response before the transformed 4xx > response is sent to the client. For what it's worth, I agree with Noel that this would be a worthwhile addition to postconf(5)'s section on soft_bounce. -- Ben Rosengart "Like all those possessing a library, Sendmail, Inc. Aurelian was aware that he was guilty of +1 718 431 3822 not knowing his in its entirety [...]" -- Jorge Luis Borges NOTICE: If received in error, please destroy and notify sender. Sender does not waive confidentiality or privilege, and use is prohibited.
Re: soft_bounce=yes in postmulti setup
On 8/9/2012 11:27 AM, Wietse Venema wrote: > It might be practical to add a note that in some cases, soft_bounce=yes > is implemented by modifying server responses. Therefore, the responses > that Postfix will log may differ from the responses that Postfix > actually sends or receives. This will be sufficient for the few > people that we write documentation for. > > Wietse > Indeed, a brief note would have been sufficient in this case. Note: postfix may log a 5xx response before the transformed 4xx response is sent to the client. -- Noel Jones
Re: soft_bounce=yes in postmulti setup
Noel Jones: > On 8/9/2012 6:51 AM, Wietse Venema wrote: > > > > > The CLEANUP DAEMON rejects the header with 5cc > > > > The SMTP DAEMON transforms the reply into 4XX. > > > > The CLEANUP DAEMON does not know that the SMTP DAEMON has soft_bounce=yes. > > > > maybe a note on the soft_bounce docs? It would not be practical to enumerate all possible instances where a "hard" error might end up being transformed into a "defer" error. With the SMTP server alone, this includes but is not limited to: - Reject in header/body checks. - Reject in smtpd_proxy content filters. - Reject in Milters. - Reject in smtpd policy lookups. With the SMTP client, there is an entirely different universe of "hard" errors that might be softened. The local and virtual delivery agents introduce their own hard error domain. It might be practical to add a note that in some cases, soft_bounce=yes is implemented by modifying server responses. Therefore, the responses that Postfix will log may differ from the responses that Postfix actually sends or receives. This will be sufficient for the few people that we write documentation for. Wietse
Re: soft_bounce=yes in postmulti setup
On 8/9/2012 6:51 AM, Wietse Venema wrote: > > The CLEANUP DAEMON rejects the header with 5cc > > The SMTP DAEMON transforms the reply into 4XX. > > The CLEANUP DAEMON does not know that the SMTP DAEMON has soft_bounce=yes. > maybe a note on the soft_bounce docs? Something like: Note: with "soft_bounce = yes", rejects generated by the cleanup daemon will be logged as 5xx, but the client will still receive a 4xx code. That at least conveys the idea that sometimes 5xx will be logged. -- Noel Jones
Re: soft_bounce=yes in postmulti setup
mailing list subscriber: [ Charset ISO-8859-1 unsupported, converting... ] > On Wed, Aug 8, 2012 at 4:19 PM, Noel Jones wrote: > > On 8/8/2012 4:22 AM, mailing list subscriber wrote: > >> Hi, > >> I asked this earlier but maybe skipped your attention. > >> > >> In summary, > >> > >> Why do body/header checks generate 5xx rejects even with soft_bounce=yes? > > > > > > soft_bounce works as documented. The log may show a 5.x.x extended > > status code, but the client receives a 4xx response and a 4.x.x status. > > > > thanks for explanation, I did not know that! > > Wietse, why would anyone want to read false information in the logs? > for example "user unknown" shows 4xx with soft_bounce=yes. are there > any others? where can I find a list of rejects that lie to me with 5xx > and I actually send a 4xx? The CLEANUP DAEMON rejects the header with 5cc The SMTP DAEMON transforms the reply into 4XX. The CLEANUP DAEMON does not know that the SMTP DAEMON has soft_bounce=yes.
Re: soft_bounce=yes in postmulti setup
On Wed, Aug 8, 2012 at 4:19 PM, Noel Jones wrote: > On 8/8/2012 4:22 AM, mailing list subscriber wrote: >> Hi, >> I asked this earlier but maybe skipped your attention. >> >> In summary, >> >> Why do body/header checks generate 5xx rejects even with soft_bounce=yes? > > > soft_bounce works as documented. The log may show a 5.x.x extended > status code, but the client receives a 4xx response and a 4.x.x status. > thanks for explanation, I did not know that! Wietse, why would anyone want to read false information in the logs? for example "user unknown" shows 4xx with soft_bounce=yes. are there any others? where can I find a list of rejects that lie to me with 5xx and I actually send a 4xx?
Re: soft_bounce=yes in postmulti setup
On 8/8/2012 4:22 AM, mailing list subscriber wrote: > Hi, > I asked this earlier but maybe skipped your attention. > > In summary, > > Why do body/header checks generate 5xx rejects even with soft_bounce=yes? soft_bounce works as documented. The log may show a 5.x.x extended status code, but the client receives a 4xx response and a 4.x.x status. -- Noel Jones
Re: soft_bounce=yes in postmulti setup
Hi, I asked this earlier but maybe skipped your attention. In summary, Why do body/header checks generate 5xx rejects even with soft_bounce=yes? Thanks. On Mon, Jul 30, 2012 at 1:21 AM, mailing list subscriber wrote: > Hi, > I have two issues. > After hard reading of MULTI_INSTANCE_README I've come to a setup as below. > I've been so far satisfied with achieving splitting postfix-2.9.3 > (ubuntu 12.04 amd64) into incoming mail, submission and delivery to > cyrus. Just before switching old box with new one, I have enabled > soft_bounce = yes. The mail delivery was interrupted for a couple of > hours, so incoming mail hit my isp backed-up mx during this time, to > which I connected and manually issued an ETRN. > > It seems however that once deliveries started, logfiles show 5xx > rejects instead of expected 4xx. Is this normal? [...]
Re: soft_bounce=yes in postmulti setup
update: interesting, header_check gives 5xx but user unknown gives correct 4xx response: Jul 30 01:49:01 mailhost postfix-in/smtpd[3685]: NOQUEUE: reject: RCPT from relay3.isp.ro[IP_ADDR]: 450 4.1.1 : Recipient address rejected: User unknown; from= to= proto=ESMTP helo=