Re: About to-addresses and followup [poll]

Re: About to-addresses and followup [poll]

[Steven L. Baur]

I'm not very happy with followup right now, and I'm including a
sample of what I've seen when doing it.  (Cut & pasted after hitting
``f'' on Sudish's message).

To: ding@ifi.uio.no
CC: The Ding list <ding@ifi.uio.no>

I'm finding the practice of cc'ing the original author along with the
mailing list increasingly irritating.  With Usenet and sprintlink (our
basic upstream feed) it's useful, since news is frequently lost.  But
with a mailing list, it just means getting two (almost, but not quite,
identical) copies of the same message.  Please enlighten me what this
practice is useful for.  I'm not so biased at this point I can't
change my mind.

> a) Add the original author to the CC list.  So we have to-address in
>    the To header, and the original author in the CC header.

> b) Add both to-address and original author to To: header.

Do it the Gnus way.  Do code for both and select by a user
configurable variable.  IMHO, you should also allow for a followup
style of just using the to-address.

How will gnus-auto-mail-to-author relate?

-- 
steve@miranova.com baur

Re: About to-addresses and followup [poll]

[Sudish Joseph]

Steven L Baur writes:
> I'm finding the practice of cc'ing the original author along with
> the mailing list increasingly irritating.  With Usenet and
> sprintlink (our basic upstream feed) it's useful, since news is
> frequently lost.  But with a mailing list, it just means getting two
> (almost, but not quite, identical) copies of the same message.
> Please enlighten me what this practice is useful for.  I'm not so
> biased at this point I can't change my mind.

For lots of mailing lists, it is common that the person who originated
the thread is not a member of the list.  For e.g., lists related to
software--a person asks a question but isn't interested in other
topics being discussed.  If we discard CC'ed addresses, this person
will lose all second and higher level messages in the ensuing
discussion.  If we don't respect even the To: header, people will get
no replies whatsoever.  I pray we don't try to enforce such a model.   

Duplicate elimination is a simple process with most filters.

> How will gnus-auto-mail-to-author relate?

It doesn't enter the picture for mailgroups.

-Sudish

PS: GNUS didn't insert a References: header on your reply, could you
see why? 

Re: About to-addresses and followup [poll]

[Steven L. Baur]

>>>>> "Sudish" == Sudish Joseph <joseph@cis.ohio-state.edu> writes:

    Sudish> PS: GNUS didn't insert a References: header on your reply,
    Sudish> could you see why?

Let me answer the somewhat obscure question first.  I traced it as far
as tm.  Once I turn off Mime, the References are inserted (as they are
in this followup).

Speaking of bugs, here's another variant of the headers on followup:

To: ding@ifi.uio.no
CC: (Steven L. Baur), The Ding List <ding@ifi.uio.no>

    Steve> (I question the rationale for the practice of Cc'ing by
    Steve> default to the author on followup to a mailing list)

    Sudish> For lots of mailing lists, it is common that the person
    Sudish> who originated the thread is not a member of the list.

I thought this was considered bad etiquette in general.  But if that
has changed, that's O.K.

    Sudish> For e.g., lists related to software--a person asks a
    Sudish> question but isn't interested in other topics being
    Sudish> discussed.  If we discard CC'ed addresses, this person
    Sudish> will lose all second and higher level messages in the
    Sudish> ensuing discussion.  If we don't respect even the To:
    Sudish> header, people will get no replies whatsoever.  I pray we
    Sudish> don't try to enforce such a model.

O.K.  It's enough for me to know there are passionate supporters for
this feature.  Be conservative in what you send, liberal in what you
accept ...

    Sudish> Duplicate elimination is a simple process with most
    Sudish> filters.

And the simplest approach (as with procmail) has its drawbacks.  What
confuses me greatly is mistaking the private copy of the public
version (either to a mailing list or on Usenet), as private e-mail.
I'll get over it.

    Steve> How will gnus-auto-mail-to-author relate?

    Sudish> It doesn't enter the picture for mailgroups.

Why shouldn't it?  Besides, if you base the behavior on a variable,
won't it be easier to get variant behavior out of the (local)
variables of the Score file, or off the group hook?

Regards,
-- 
steve@miranova.com baur

Re: About to-addresses and followup [poll]

[Sudish Joseph]

Steven L Baur writes:
>>>>>> "Sudish" == Sudish Joseph <joseph@cis.ohio-state.edu> writes:
Sudish> Duplicate elimination is a simple process with most
Sudish> filters.

> And the simplest approach (as with procmail) has its drawbacks.  What
> confuses me greatly is mistaking the private copy of the public
> version (either to a mailing list or on Usenet), as private e-mail.
> I'll get over it.

There's no such thing as a "private/public" copy of duplicated email.
Each copy is the same as the other, barring Received: and any other
headers added in-transit.  Both copies contain exactly the same
information, and have the same message-id.  This is not true of
newsgroups and mailed copies of postings.

Steve> How will gnus-auto-mail-to-author relate?
Sudish> It doesn't enter the picture for mailgroups.

> Why shouldn't it?  Besides, if you base the behavior on a variable,
> won't it be easier to get variant behavior out of the (local)
> variables of the Score file, or off the group hook?

gnus-followup-to-function exists for this very purpose.  

-Sudish

Re: About to-addresses and followup [poll]

[Per Abrahamsen]

>>>>> "Sudish" == Sudish Joseph <joseph@cis.ohio-state.edu> writes:

Sudish> For lots of mailing lists, it is common that the person
Sudish> who originated the thread is not a member of the list.

>>>>> "SLB" == Steven L Baur <steve@miranova.com> writes:

SLB> I thought this was considered bad etiquette in general.  But if that
SLB> has changed, that's O.K.

It hasn't changed.  Good etiquette has always mandated that you
followed the policies for a particular mailing list, and that the
policies were determined by the list owner.  

Examples of mailing lists where messages from non-subscribers are
welcome and encouraged are auc-tex, all the GNU bug-* and help-*
lists, and lists like `sales@acme.com'.

Exmaples where messages from non-subscribers are not welcome are are
most newsgroups, the gcc2 list for gcc developers, and perhaps the
ding list (messages from non-subscribers should probably use the gnus
or gnus-bug lists instead).

Re: About to-addresses and followup [poll]

[Sten Drescher]

In list.ding, Sudish Joseph <joseph@cis.ohio-state.edu> said:

SJ> Steven L Baur writes:

>> How will gnus-auto-mail-to-author relate?

SJ> It doesn't enter the picture for mailgroups.

	It doesn't?  Then why did Gnus CC: both Steve Baur (from the To:
header) and you (from the From: header) (before I edited the header to
remove Steve)?

-- 
#include <disclaimer.h>				/* Sten Drescher */
To get my PGP public key, send me email with your public key and
	Subject: PGP key exchange
Key fingerprint =  90 5F 1D FD A6 7C 84 5E  A9 D3 90 16 B2 44 C4 F3

Re: About to-addresses and followup [poll]

[Steven L. Baur]

>>>>> "Per" == Per Abrahamsen <abraham@dina.kvl.dk> writes:
>>>>> "Sudish" == Sudish Joseph <joseph@cis.ohio-state.edu> writes:

    Sudish> For lots of mailing lists, it is common that the person
    Sudish> who originated the thread is not a member of the list.

I got a personal example of the good intent of this yesterday evening,
when a question I asked in a *-user list got responded to in the
corresponding *-worker list.  Point taken, even though I subscribe to
the *-worker list.

>>>>> "SLB" == Steven L Baur <steve@miranova.com> writes:

    SLB> I thought this was considered bad etiquette in general.  But
    SLB> if that has changed, that's O.K.

    Per> It hasn't changed.  Good etiquette has always mandated that
    Per> you followed the policies for a particular mailing list, and
    Per> that the policies were determined by the list owner.

That seems to me to cry out for something to set out of the group
parameters, and/or in the active file.

-- 
steve@miranova.com baur

Re: About to-addresses and followup [poll]

[Sudish Joseph]

Steven L Baur writes:
> That seems to me to cry out for something to set out of the group
> parameters, and/or in the active file.

I think it's important that ding should offer defaults that aren't
radically different from existing practice.  Having options is
important too, of course--but the defaults should be simple and easy
to understand.  

In this case, if the user adds a to-address to the group params, I'd
suggest that the only immediate change in behaviour be to supply a
default for `a' in that group.  To activate other options, those
should be turned on explicitly.  Modifying both `a' and `f' on the
basis of one entry in the param list is confusing.  Requiring the user
to add another entry in the param list to turn off special behaviour
on `f' (leaving any magic on `a' enabled) seems silly.

-Sudish

Re: About to-addresses and followup [poll]

[Sudish Joseph]

Sten Drescher writes:
> In list.ding, Sudish Joseph <joseph@cis.ohio-state.edu> said:
SJ> Steven L Baur writes:
>>> How will gnus-auto-mail-to-author relate?
SJ> It doesn't enter the picture for mailgroups.
> 	It doesn't?  Then why did Gnus CC: both Steve Baur (from the To:
> header) and you (from the From: header) (before I edited the header to
> remove Steve)?

Traditionally, mail agents have provided (at least) 2 forms of the
"reply" function with respect to who gets copies of the reply.  

The first form is where you wish to reply to only the sender(s) of the
message, ignoring all others who received a copy of the same message.
GNUS binds this action to `r' in the summary.

The second form is where you reply to everyone who received a copy of
the original message + the sender(s).  GNUS binds this to `f' in the
summary. 

Every MUA I've seen supports this distinction.  The important thing to
note here is that the sender always gets a copy of the reply, unless
the user explicitly deletes her address.  GNUS breaks an important
element in this with it's behaviour w.r.t. a user pressing `f' in a
group where to-address is defined--everyone _except_ the sender
receives a copy.  Very weird and confusing, IMO.

gnus-auto-mail-to-author fills a different need.  Unlike in mail (or
mailing lists, which are _NOT_ newsgroups, though we're lucky enough
to have a common interface to both), netiquette demands that you read
the group that you post a question to.  In such a situation, it is
superfluous to email a copy of the article you've posted back to the
author.  In some cases, it is useful to CC: a copy--unreliable/slow
newsfeeds, faster transmission, etc.  This is where
gnus-auto-mail-to-author comes in.

Or you could 'agrep -d defun gnus-auto-mail-to-author gnus-msg.el' and
see that it only occurs in functions that post to USENET.  :)

-Sudish

Re: About to-addresses and followup [poll]

[Lars Magne Ingebrigtsen]

So, gathering these suggestions, I think we need three, uhm, thingies:

1)  Ordinary mail groups
When you press, `f', the From is put in the To, and the Cc and To is
put in the Cc.

2) Closed mailing list groups.
These are lists that only subscribers post to.  The user will add a
`to-address' parameter, and `a' and `f' will use that address and that
address only -- all From, To and Cc headers are ignored.

3) Open mailing list groups.
The user will add a `to-list' group parameter.  This is used when
doing the `a' command.  When doing an `f', the From is put in the To,
and the To and Cc are put in the Cc.  

1) and 2) are what is in Gnus 5 while 3) is new.  I think this should
cover everything.  Or most common cases, anyway.

-- 
Home is where the cat is.

Re: About to-addresses and followup [poll]

[Steven L. Baur]

>>>>> "Lars" == Lars Magne Ingebrigtsen <larsi@ifi.uio.no> writes:

    Lars> So, gathering these suggestions, I think we need three, uhm,
    Lars> thingies:

    Lars> 1) and 2) are what is in Gnus 5 while 3) is new.  I think
    Lars> this should cover everything.  Or most common cases, anyway.

I don't suppose you could throw in a Mail-Copies-To: while you're at
it?

I'll take a ``Mail-Copies-To: never'' to go, please :-)

-- 
steve@miranova.com baur

Re: About to-addresses and followup [poll]

[Per Abrahamsen]

>>>>> "LMI" == Lars Magne Ingebrigtsen <larsi@ifi.uio.no> writes:

LMI> 2) Closed mailing list groups.
LMI> These are lists that only subscribers post to.  The user will add a
LMI> `to-address' parameter, and `a' and `f' will use that address and that
LMI> address only -- all From, To and Cc headers are ignored.

Well, if the people on the CC is there for a reason, then they are not
on the mailing list.  Thus, I don't think the To and CC fields should
be ignored.

Re: About to-addresses and followup [poll]

[Lars Magne Ingebrigtsen]

steve@miranova.com (Steven L. Baur) writes:

> I don't suppose you could throw in a Mail-Copies-To: while you're at
> it?

Well, why not?  

> I'll take a ``Mail-Copies-To: never'' to go, please :-)

This means that if you put that line in all mail you send to the ding
mailing list, you'll never get those pesky double messages (from
people using September Gnus 0.15 and up.)  :-)

-- 
Home is where the cat is.

Re: About to-addresses and followup [poll]

[Lars Magne Ingebrigtsen]

Per Abrahamsen <abraham@dina.kvl.dk> writes:

> Well, if the people on the CC is there for a reason, then they are not
> on the mailing list.  Thus, I don't think the To and CC fields should
> be ignored.

Well, if it's reasonable to assume that everybody that posts to a
mailing list also reads it (which is the case on this list), then
there should be a way to say "never to any Cc'ing".  Which is what
`to-address' does.

For instance, this is from the head of your mail:

> From: Per Abrahamsen <abraham@dina.kvl.dk>
> To: larsi@ifi.uio.no (Lars Magne Ingebrigtsen)
> CC: ding@ifi.uio.no

Standard `f' would have a mail go to me, you and ding, I think.  But I
really want to send just to ding.

-- 
Home is where the cat is.

Re: About to-addresses and followup [poll]

[Felix Lee]

> Standard `f' would have a mail go to me, you and ding, I think.  But I
> really want to send just to ding.

but this doesn't do the right thing.  sometimes I get mail on lists
that involves people who aren't on the list, who in turn reply to the
list.  unless you know who's on the mailing list, you can't delete
addresses safely.
--

Re: About to-addresses and followup [poll]

[Lars Magne Ingebrigtsen]

Felix Lee <flee@teleport.com> writes:

> > Standard `f' would have a mail go to me, you and ding, I think.  But I
> > really want to send just to ding.
> 
> but this doesn't do the right thing.  sometimes I get mail on lists
> that involves people who aren't on the list, who in turn reply to the
> list.  unless you know who's on the mailing list, you can't delete
> addresses safely.

That's true, but it is very convenient.  For instance, whenever I do
an `f' in this group, I get just "ding@ifi.uio.no" in the To header,
and that makes me very happy.

-- 
Home is where the cat is.

Re: About to-addresses and followup [poll]

[Sudish Joseph]

Lars Magne Ingebrigtsen writes:
> Felix Lee <flee@teleport.com> writes:
> That's true, but it is very convenient.  For instance, whenever I do
> an `f' in this group, I get just "ding@ifi.uio.no" in the To header,
> and that makes me very happy.

I think it'd be nice if the interface to this stuff was changed
slightly.  Why not have just one parameter, to-address, that specifies
the address to be used, and a variable or group parameter that selects
the desired behaviour (use to-address only for `a'/use it to clobber
everything else)?  This would make it easier to customise, too.  

The current use of two different parameters, both of which specify the
same type of data is confusing.  At the very least, exchange the
sematics of the to-list and to-address.

-Sudish

Re: About to-addresses and followup [poll]

[Lars Magne Ingebrigtsen]

Sudish Joseph <joseph@cis.ohio-state.edu> writes:

> I think it'd be nice if the interface to this stuff was changed
> slightly.  Why not have just one parameter, to-address, that specifies
> the address to be used, and a variable or group parameter that selects
> the desired behaviour (use to-address only for `a'/use it to clobber
> everything else)?  This would make it easier to customise, too.  

A group parameter and a variable to customize the effect?  I think not
-- I currently use both `to-list' and `to-address', so both are useful
in different groups.

> The current use of two different parameters, both of which specify the
> same type of data is confusing.  At the very least, exchange the
> sematics of the to-list and to-address.

That would be incompatible with Gnus 5.1, so it's too late for
something like that.

-- 
(domestic pets only, the antidote for overdose, milk.)
  larsi@ifi.uio.no * Lars Ingebrigtsen

Re: About to-addresses and followup [poll]

[Sudish Joseph]

Per Abrahamsen writes:
>>>>>> "SJ" == Sudish Joseph <joseph@cis.ohio-state.edu> writes:
SJ> If a list is closed in the sense that only authorized people can
SJ> post to it, this whole issue is irrelevant, coz the Sender's
SJ> message will not reach the list.

> Huh?  That doesn't make any sense to me.

Probably because we seem to be discussing different issues. :-) My
only worry here is who receives a copy of any replies made to a
message--eliminating the sender seems cruel to me when CC'ed addresses
aren't.  Yours seems to be to eliminate duplicates for closed lists.
I'd assumed that you were refering to the problem of non-subscribers
receiving copies of messages they shouldn't.  So, read the above like
this: "Since only subscribers may post to the list, messages from
non-subscribers will not reach the list.  In which case, dropping the
From: address in replies will not serve any purpose whatsoever."
(The "purpose" there refers to confidentiality, not duplicate
elimination.) 


> If you reply to a closed list, we can assume both the original sender
> and you subscribe to the list.  With the `to-list' semantics, the
> original sender will receive two copies of the reply, one through the
> list and one directly from you.  With the `to-address' semtics, the
> original sender will only receive one copy. 

I confess that I'm not too excited about having to deal with two
separate options just so that people on closed lists will not receive
duplicates.  If the issue were confidentiality, I'd agree...

-Sudish

Re: About to-addresses and followup [poll]

[Per Abrahamsen]

>>>>> "SJ" == Sudish Joseph <joseph@cis.ohio-state.edu> writes:

SJ> Actually, I'd submit that the present behaviour is the opposite of
SJ> correct behaviour for closed lists.  

I defined closed two messages back in the thread.

SJ> If a list is closed in the sense
SJ> that only authorized people can post to it, this whole issue is
SJ> irrelevant, coz the Sender's message will not reach the list.  

Huh?  That doesn't make any sense to me.

If you reply to a closed list, we can assume both the original sender
and you subscribe to the list.  With the `to-list' semantics, the
original sender will receive two copies of the reply, one through the
list and one directly from you.  With the `to-address' semtics, the
original sender will only receive one copy. 

SJ> In addition, I'd expect
SJ> that such lists would definitely set Reply-To: in all mailings.

That would really be bad, since you couldn't make a private reply
then. 

SJ> Only use to-address in selecting a To: header for fresh
SJ> posts to that list (via `a' or `m').

Not via `m'!  Pushing `m' doesn't post an article in a news group and
shouldn't mail to the list in a mail group.  That would break the
symmetry between mailing lists and newsgroups.

Re: About to-addresses and followup [poll]

[Sudish Joseph]

Per Abrahamsen writes:
>>>>>> "SJ" == Sudish Joseph <joseph@cis.ohio-state.edu> writes:
SJ> But it seems redundant to have both a to-address and a
SJ> to-list parameter.

> The current behavior is better for closed list.  For example, if I had
> set to-address for the ding folder, you wouldn't receive two copies of
> this message, as you will now.

Actually, I'd submit that the present behaviour is the opposite of
correct behaviour for closed lists.  If a list is closed in the sense
that only authorized people can post to it, this whole issue is
irrelevant, coz the Sender's message will not reach the list.  If the
list is closed in the sense that only people on the list may receive
the message, the actual desired behaviour is to not include originally
CC'ed addresses in your reply; by all rules of netiquette the original
sender should receive a reply.

> As I see it, to-address should be set on closed lists and to-list
> should be set on open lists.  

I don't think GNUS should confuse this by adding another parameter.
Closed lists are just a very special case.  In addition, I'd expect
that such lists would definitely set Reply-To: in all mailings.

I think your original suggestion is much better.  Don't use to-address
(or any other parameter) when following up.  The old behaviour worked
just fine--since you received it from the list, the list name will be
there in the headers (in all except a vanishingly small set of cases,
anyway).  Only use to-address in selecting a To: header for fresh
posts to that list (via `a' or `m').

This would also avoid the possibility of sending mail to a list when
you didn't want to in the first place (the "B m" scenario I outlined
elsewhere).

Mailing lists aren't the same as newsgroups, insofar as the poster's
expectations of the prospective audience is.  Or in how she expects to
receive replies (in a public forum/via personal mail).  Why treat them
in the same manner?

-Sudish

Re: About to-addresses and followup [poll]

[Per Abrahamsen]

>>>>> "SJ" == Sudish Joseph <joseph@cis.ohio-state.edu> writes:

SJ> But it seems redundant to have both a to-address and a
SJ> to-list parameter.

The current behavior is better for closed list.  For example, if I had
set to-address for the ding folder, you wouldn't receive two copies of
this message, as you will now.

As I see it, to-address should be set on closed lists and to-list
should be set on open lists.  

If the message you reply to have the `Mail-Copies-To:' header
suggested by Lars, that should be able to overwrite the default
algorithm for that folder.  Then you could set `Mail-Copies-To: never'
when writing to a list you subscribe to, and `Mail-Copies-To:
your@address' when writing to a list you don't subscribe to.

Re: About to-addresses and followup [poll]

[Sudish Joseph]

Per Abrahamsen writes:
> I believe sgnus have (or will have) a to-list parameter which is only
> used on new posts (i.e. when you type `a') and not on followups.  The
> default followup algorithm will send to both the list and the author
> if there is no to-address on the list.  The author in the To: header
> and the list in the CC: header, as in this message.

I agree with you that it's much better to have the list address in the
CC: field than in the To: field (the headers on this message look a
tad ugly :-).  But it seems redundant to have both a to-address and a
to-list parameter.  We can achieve the same functionality by simply
ignoring to-address on followups, limiting its scope to just new
posts.

Alternatively, to avoid the Pineism I mentioned in another message, we
might retain the current meaning of to-address (ie., use it on
followups) and prompt the user if we detect a followup to a mail that
didn't have (mail-strip-quoted-names to-address) in it's original
headers ("This message wasn't send to" to-address ", should we Cc that
address?").  This would cover all possibilities, I think?

-Sudish

Re: About to-addresses and followup [poll]

[Per Abrahamsen]

The current behavior (no copy to the author on followups) make sense
on closed lists, i.e. lists where only subscribers may post.  I
suspect most lists are open though.  All the gnu lists are open, for
example.

I believe sgnus have (or will have) a to-list parameter which is only
used on new posts (i.e. when you type `a') and not on followups.  The
default followup algorithm will send to both the list and the author
if there is no to-address on the list.  The author in the To: header
and the list in the CC: header, as in this message.

About to-addresses and followup [poll]

[Sudish Joseph]

Quick poll, which of these two do you prefer?

Background:  If you have to-address set in a mail group, and you do a
followup, the original author isn't CC'ed a copy.  Instead, the
to-address replaces the authors name.  Note that the people who were 
CC'ed still get a copy, only the original author doesn't.

a) Add the original author to the CC list.  So we have to-address in
   the To header, and the original author in the CC header.
b) Add both to-address and original author to To: header.

I have a patch to fix this, it does (a) above.  But it struck me that
given the original meaning of To: and what other mailers do on a
followup, it makes more sense to do (b).  This might be a cosmetic
issue, but it would irk those who do differentiate on whether they
were to'ed or cc'ed a given message.

Opinions?

-Sudish