On 4/30/20 7:12 PM, Joel M. Halpern wrote:
Keith, why do you say that last call comments are going all over the
place. The last call announcement sets reply-to to the last-call
list. So unless people deliberately change it, that is where those
comments go.
I haven't tried to survey MUAs for this behavior in a few years, but in
practice I don't think you can depend on reply-to to be treated
consistently by all MUAs.
The behavior of email users, MUAs, list software, etc. are all somewhat
unpredictable, and I suspect inherently so, for a variety of reasons
which for the sake of brevity I'll not mention here. So it's difficult
for the sender of a message to force recipients to behave in the ways
that the sender specifies.
-----
The Last Call announcements go to lost of places: ietf-announce, the
working group mailing list, a specific email address for the document
(draft-xxx-yyy-zzz@xxxxxxxx), {wg}-chairs@xxxxxxxx, [*] and separately,
the addresses of the chairs of the WG.
The reply-to on the Last Call announcement is indeed set to
last-call@xxxxxxxx. However the mailing list typically adds a
List-Post header, and some MUAs encourage replies to go to that header
when it is present.
IMO it might be desirable for a WG mailing list to be in the loop for
Last Call comments (and for the WG list to be cc'ed on announcements),
but I would still like all of those comments to be collected together
and made publicly visible in one obvious place. Whether the best place
for that is the last-call@xxxxxxxx list, or the datatracker, or some
other place, I'm not entirely sure. For archive purposes the
datatracker seems good, for the purpose of having an effective ongoing
conversation there's an advantage to keeping email messages in the email
system and using ordinary email tools to deal with them.
The text of the announcement does say "Please send substantive comments
to the last-call@xxxxxxxx mailing lists [sic] by {date}". So that's good.
-----
I'm tempted to suggest that every Last Call announcement should be From
and have Reply-To set to last-call-{number}@ietf.org (where {number} is
unique to that Last Call) and for that address to forward the message
(adding/modifying To/Cc header fields as necessary) to wherever it needs
to go: perhaps last-call@xxxxxxxx, perhaps the WG mailing list, perhaps
the datatracker, certainly an archive. The same
last-call-{number}@ietf.org should be specified in the announcement text
itself. That way, all of the responses would be funneled through one
place and processing/archiving/etc. would be consistent. Searching of
messages would also be more reliable as one could look for that specific
address in the to/cc message headers.
Adding {area} and {wg}, e.g. last-call-{area}-{wg}-{number}@ietf.org,
where applicable, might facilitate easy searching for Last Call traffic
in a particular area or WG.
What makes me only "tempted" to suggest this, is that I don't know
whether people have email filters in place that this would confuse (lots
of people expect intended recipient addresses to appear in the to/cc
headers, even though this has never been required or expected), or if
there are still some WG lists hosted outside of ietf.org that would
alter the reply-to field, etc. The email network is a vast and diverse
place.
For the moment, these are the recommendations that make sense to me as
being safe changes:
1. Announcements should CC last-call@xxxxxxxx in addition to the long
list of other CCs.
2. Last Call comments submitted via the datatracker should use the same
Subject format as replies to the announcement:
Re: Last Call: <{internet-draft-identifier}> ({internet-draft-title}) to
{intended-status}
(even if the Last Call comment is generated before the announcement -
the Re: prefix is important to distinguish the responses from the
announcement itself)
3. The Last Call responses, along with the announcement, should be made
easily viewable in some way, perhaps via the datatracker, perhaps via an
email archive that's specific to the Last Call.
(I've not found the IETF email archives to be very usable in the past,
but improving them might make more sense than fixing the datatracker to
be a good email archive.)
Keith
[*] notation: I'm using curly braces {} as a substitute for italics,
because some email processors don't preserve HTML and some of them
translate HTML in weird ways that make it less readable than plain text.
