Thread: Re: reply-to set

Re: reply-to set

From
Josh Berkus
Date:
(crossing over to pgsql-www, because I think we need to hash out whether
this is a good idea or not)

On 07/30/2013 10:07 AM, Alvaro Herrera wrote:> Josh Berkus wrote:
>> On 07/30/2013 09:40 AM, Alvaro Herrera wrote:
>>> Hi,
>>>
>>> I have set the reply_to option for lists pgsql-jobs and pgsql-announce
>>> to $SENDER, which (if I read the manual right) means that the outgoing
>>> messages will have a Reply-To: header pointing to the sender of the
>>> message.  The intent is to prevent some mistakes where people reply to
>>> those lists instead of to the senders.
>>>
>>> I hope this will work sanely, but if anything looks wrong to you, please
>>> ping me immediately.  Thanks.
>>
>> Did we discuss this on -www?
>
> No, only in the sysadmins channel.
>
>> That makes sense for -jobs, but I'm not sure it's such a good idea for
>> -announce.
>
> Why not?  Surely replies to -announce shouldn't go to the list.  If you
> have another idea to set the reply-to for pgsql-announce, I'm all ears.
>

People who post to -announce already get dozens of bounce messages as it
is.  If you do reply-to-sender:

a) the number of bounce messages an -announce poster gets will go into
the hundreds (as used to be the case)

b) majordomo won't do automated bounce processing if bounces don't hit
the list, so invalid subscribers will never be removed.

> Surely replies to -announce shouldn't go to the list.  If you
> have another idea to set the reply-to for pgsql-announce, I'm all ears.

Ideally, we'd have an address which would trigger automated bounce
processing without ending up in list moderation if it's not treated as a
bounce.  No idea how to make that work with MJ, though.


-- 
Josh Berkus
PostgreSQL Experts Inc.
http://pgexperts.com



Re: reply-to set

From
Greg Stark
Date:
On Tue, Jul 30, 2013 at 6:31 PM, Josh Berkus <josh@agliodbs.com> wrote:
> a) the number of bounce messages an -announce poster gets will go into
> the hundreds (as used to be the case)
>
> b) majordomo won't do automated bounce processing if bounces don't hit
> the list, so invalid subscribers will never be removed.

Bounces do not follow reply-to, they go to the envelope sender. I'm
assuming MJ does VRP which sets the envelope sender to a unique
address for each message so it can process the bounce accurately.
Mailer that sends a bounce to the reply-to are the same broken ones
that would send it to the author in the From header if reply-to isn't
set and wouldn't be possible to process automatically in any case.

I'm normally against setting reply-to. In these two cases it would be
defensible.  The main problem I see is that users who send mail to
these lists probably should be already setting reply-to anyways
themselves. They probably don't want to receive personal responses but
to have replies go to their own internal customer support lists. I
assume MK keeps the reply-to if it's already present?


-- 
greg



Re: reply-to set

From
Alvaro Herrera
Date:
Josh Berkus wrote:

> > Why not?  Surely replies to -announce shouldn't go to the list.  If you
> > have another idea to set the reply-to for pgsql-announce, I'm all ears.
> 
> People who post to -announce already get dozens of bounce messages as it
> is.  If you do reply-to-sender:
> 
> a) the number of bounce messages an -announce poster gets will go into
> the hundreds (as used to be the case)
> 
> b) majordomo won't do automated bounce processing if bounces don't hit
> the list, so invalid subscribers will never be removed.

No, bounces (should) go to the Sender address, not the reply-to address;
and the Sender is still set to the mj2 address which passes it to the
bounce processor and removes subscribers that bounce too much.

I haven't ever posted to pgsql-announce so I don't know how many bounces
do they get, but I would like to know.  Really, they shouldn't get ANY
bounce at all; and if they do, they should notify the mj2 admin (me)
about them.

> > Surely replies to -announce shouldn't go to the list.  If you
> > have another idea to set the reply-to for pgsql-announce, I'm all ears.
> 
> Ideally, we'd have an address which would trigger automated bounce
> processing without ending up in list moderation if it's not treated as a
> bounce.  No idea how to make that work with MJ, though.

That already works.  Invalid subscribers have been removed in dozens
since I fixed the list config some weeks ago.  (Several hundred
invalid addresses were removed from pgsql-announce the first time the
threshold was crossed ... about 4 weeks ago, I think.)

If you have problems with a list you administer, let me know.

-- 
Álvaro Herrera                http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Training & Services



Re: reply-to set

From
Alvaro Herrera
Date:
Greg Stark wrote:

> I'm normally against setting reply-to. In these two cases it would be
> defensible.  The main problem I see is that users who send mail to
> these lists probably should be already setting reply-to anyways
> themselves. They probably don't want to receive personal responses but
> to have replies go to their own internal customer support lists. I
> assume MK keeps the reply-to if it's already present?

Yes -- if there's a reply-to already, it's not changed.

-- 
Álvaro Herrera                http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Training & Services



Re: reply-to set

From
Josh Berkus
Date:
Alvaro,


> No, bounces (should) go to the Sender address, not the reply-to address;
> and the Sender is still set to the mj2 address which passes it to the
> bounce processor and removes subscribers that bounce too much.
> 
> I haven't ever posted to pgsql-announce so I don't know how many bounces
> do they get, but I would like to know.  Really, they shouldn't get ANY
> bounce at all; and if they do, they should notify the mj2 admin (me)
> about them.

There are always bad mail hosts who send any bounce message to FROM (or
REPLY-TO) instead of any other address.  For example, we've regexed out
Brazillian host UOL.com from the PostgreSQL mailing lists because (a)
they require manual sender confirmation, and (b) the confirmation
message goes to the FROM address, not SENDER or REPLY-TO.

Right now, though, it's only 10-20 per annnouncement, which is
tolerable.  Heck, it's generally my confirmation that the annoucement
went out.

> That already works.  Invalid subscribers have been removed in dozens
> since I fixed the list config some weeks ago.  (Several hundred
> invalid addresses were removed from pgsql-announce the first time the
> threshold was crossed ... about 4 weeks ago, I think.)

Oh, great!

In general, I think reply-to for announce should be set to a black hole
address (e.g. do-not-reply@postgresql.org).  There's never a good reason
to reply to an -announce message.

-- 
Josh Berkus
PostgreSQL Experts Inc.
http://pgexperts.com



Re: reply-to set

From
Greg Stark
Date:
On Tue, Jul 30, 2013 at 6:52 PM, Josh Berkus <josh@agliodbs.com> wrote:
> There are always bad mail hosts who send any bounce message to FROM (or
> REPLY-TO) instead of any other address.

Yes, but setting reply-to won't make those any worse. We should really
ban any such hosts from our lists.

> In general, I think reply-to for announce should be set to a black hole
> address (e.g. do-not-reply@postgresql.org).  There's never a good reason
> to reply to an -announce message.

I disagree with noreply addresses on principle. They usually just
represent the sender being lazy and not thinking hard enough about
where replies should go. Why wouldn't there be a good reason to reply
to an announcement? I've often wanted to reply to announcements.


-- 
greg



Re: reply-to set

From
Josh Berkus
Date:
> I disagree with noreply addresses on principle. They usually just
> represent the sender being lazy and not thinking hard enough about
> where replies should go. Why wouldn't there be a good reason to reply
> to an announcement? I've often wanted to reply to announcements.

Because the poster of the annoucement may be largely unrelated to its
content.  Often the address to which any comments need to be addresses
is somewhere in the announcement text.  The actual poster of the
announcement may be a community member posting on behalf of someone
else, or a PR company posting from an intern's address.

For a PostgreSQL release announcement, for example, we want anybody who
has a comment or question to send mail to press@postgresql.org, NOT to
the sysadmin who posted the actual -announce email.  Having a noreply@
mailing address would enforce that.

-- 
Josh Berkus
PostgreSQL Experts Inc.
http://pgexperts.com



Re: reply-to set

From
Greg Stark
Date:
<p dir="ltr"><br /> On Jul 30, 2013 8:24 PM, "Josh Berkus" <<a
href="mailto:josh@agliodbs.com">josh@agliodbs.com</a>>wrote:<br /> ><br /> ><br /> > > I disagree with
noreplyaddresses on principle. They usually just<br /> > > represent the sender being lazy and not thinking hard
enoughabout<br /> > > where replies should go. Why wouldn't there be a good reason to reply<br /> > > to an
announcement?I've often wanted to reply to announcements.<br /> ><br /> > Because the poster of the annoucement
maybe largely unrelated to its<br /> > content. <p dir="ltr">That's what reply-to is for.<p dir="ltr">><br />
>For a PostgreSQL release announcement, for example, we want anybody who<br /> > has a comment or question to
sendmail to <a href="mailto:press@postgresql.org">press@postgresql.org</a>, NOT to<br /> > the sysadmin who posted
theactual -announce email.  Having a noreply@<br /> > mailing address would enforce that.<p dir="ltr">It would
enforceit by totally breaking replies unless the user took manual intervention.  

Re: reply-to set

From
Alvaro Herrera
Date:
Josh Berkus wrote:

> For a PostgreSQL release announcement, for example, we want anybody who
> has a comment or question to send mail to press@postgresql.org, NOT to
> the sysadmin who posted the actual -announce email.  Having a noreply@
> mailing address would enforce that.

So what you really want is for announcements to carry
a Reply-To: press@postgresql.org header, which you can do already and
wouldn't be broken by the change I'm introducing.

-- 
Álvaro Herrera                http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Training & Services



Re: reply-to set

From
Josh Berkus
Date:
On 07/30/2013 03:08 PM, Alvaro Herrera wrote:
> Josh Berkus wrote:
> 
>> For a PostgreSQL release announcement, for example, we want anybody who
>> has a comment or question to send mail to press@postgresql.org, NOT to
>> the sysadmin who posted the actual -announce email.  Having a noreply@
>> mailing address would enforce that.
> 
> So what you really want is for announcements to carry
> a Reply-To: press@postgresql.org header, which you can do already and
> wouldn't be broken by the change I'm introducing.

OK, makes sense.  We'll need to promulgate some information for posters
to -announce though on how to set Reply-To.


-- 
Josh Berkus
PostgreSQL Experts Inc.
http://pgexperts.com