Re: mail-{followup,copies}-to on this list

[Per Abrahamsen <abraham@dina.kvl.dk>]

NAGY Andras <nagya@inf.elte.hu> writes:

> (http://www.newsreaders.com/misc/mail-copies-to.html) says
> 
>   If the header is not present the agent SHOULD NOT set the message to
>   be both posted and mailed.
> 
> Please correct me if I interpret this wrong, or stop sending me
> personal followups.

The document is irrelevant, as the document in question only covers
the use of the header on Usenet, not email.  The situation for email
are very different than Usenet, both for cultural and technical
reasons.

However, while the issue is not covered by any standards, there is the
much more important question of politeness.  

Of course, in general it is silly to send multiple copies of the same
message by different paths to the same receiver.  Unfortunately, when
only using the common headers, there is no machanical way an agent can
determine whether the receivers of foo@bar.org is a subset of the
receivers of baz@foobar.net.  Or more specifically, without more
information there is no way Gnus can know that nagya@inf.elte.hu is a
member of the mailing list ding@gnus.org.  To be safe, the default
action have to be to send to both addresses.

There are several solutions to this problem.  Here are some of the
ones the sender can use:

1. He can setup Gnus to sort away duplicates.  This is the one I use,
and I haven't been bothered by duplicates since that.

2. He can use "mail-copies-to: nobody", in order to tell the receiver
that he already read one of the to or cc addresses, so it can ignore
the from address.

3. He can use "mail-followup-to: list@address", to indicate which of
the adresses in question that is most appropriate for wide replies.

4. He can try flaming on the list, and hope the receivers will make it
their problem.  I strongly discourage selecting this option.

The reciecer can also do something, even if he has fewer options.

1. IF the list is closed, i.e. only subscribers send to it, and IF he
read it from a Gnus folder, and IF there are no crossmailed messages
(i.e. messages mailed to multiple lists), he can set the 'to-address'
group parameter.  I do this for some lists, e.g. ding, but not others
like auctex (which is open for non-subscribers) or xemacs-beta (which
has lots of crossmailings).

2. The can manually for each message try to guess who is on which
lists.  I do not recommend this, it is way more unreliable and takes a
lot more work for the receiver, than it would take for the sender to
solve the problem from his side.