[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Q: List Policy

Steve Lamb (2008-11-22 04:40 -0800) wrote:

> On Saturday 22 November 2008 04:15:42 Henrique de Moraes Holschuh wrote:
>> Actually, to be very blunt: CCing people is absolutely the only way
>> to deal with massive ammounts of email and very-high-traffic lists
>> when you *care* about not ignoring email that you should have read.
>     That is absolute, 100% pure rubbish. This is solvable by technical
> means, right now, today, if email client authors would just implement
> a feature [...]

I think that "being solvable" is not an option. Too many if's in your
message. I'd like to remind everybody that email is a distributed
system. We can't control what others do, we can only choose what we do
ourselves and which mail messages we pay attention to. The fact is that
mail is read with many different devices with different software and
with different features and configurations. Obviously it is also read by
different people with varying level of expertise. Mailing lists are read
through mail-to-news gateways which may or may not support bidirectional
message delivery. Some people read mail trough web-based mailing list
archives because they're only occasionally interesting in the list
discussion. Some people filter their mail differently depending on if
they are in the To field or in the Cc field. Some discussion threads are
cross-posted to two or more mailing lists because it concerns more than
one developer group. Perhaps the discussion is also CCed to some
email-based bug tracking system and the original bug reporter.

In short, we don't know how others receive, read and compose their mail,
or who are subscribers of certain mailing list (people join and leave
all the time). What kind of reply policies and email-client
configurations we should enforce for these varying situations? I think
they would soon became quite complicated. How do we make people to
understand and follow such policies?

My opinion: there's no way. Surely there are policies and there are
email standards. But there is also the practice.

On a large scale, perhaps with a little exaggeration, there are only two
quite reliable features in email:

     1. Reply
     2. Reply to all

Anything that requires more advanced technique and configuration than
these can't be trusted to work reliably. System must be based on these
two. Other features may _sometimes_ work.

People can (and very likely will) talk endlessly about the correct and
incorrect usage of Mail-Followup-To and Reply-To headers, about good and
bad email clients, good and bad configuration etc., but I don't think it
will ever prevent mail from going sometimes to wrong places. We can't
really control how other people send their mail; we can only choose what
we read (or ignore) and how we configure our own system.

So I assume that, for large audience, only "Reply" and "Reply to all"
work reliably. Based on that there are generally two ways for making a
mailing list sort of work for normal people without endless and
error-prone configuration and header-editing hassle:

     1. Tell people to press the "Reply" button and configure mailing
        list software to add Reply-To header which points to the list
     2. Tell people to press the "Reply to all" button.

Both have their advantages and disadvantages and the subject has been
pretty much discussed to death. In the option 2 a person may perhaps
edit the recipient fields manually but there's no guarantee that he will
edit it "correctly" from some other person's point of view.

Then there's the Debian way: Reply-To is not pointing to the list
address and using "Reply to all" is discouraged. Some people like this
policy. Nevertheless, it causes some difficulties: people sometimes
press "Reply" and thus send mail to the author only while expecting it
to go to the list. Sometimes they expect something else. Sometimes
people press "Reply to all" and annoy some other people with
carbon-copies and duplicate messages. So even with the Debian way,
depending on the point of view, mail sometimes goes to "wrong" places.

This is the reality and it's pretty complicated. It's not because people
are "stupid" or something; it's because it is impossible on enforce a
perfectly unified policy and client configuration in distributed system.
So let's just configure our own email clients so that dealing with the
reality and different lists is as easy as possible.

Reply to: