Bug in mailing lists; unfriendly to non-subscribers

Cameron Simpson cs at zip.com.au
Tue Jul 6 23:07:52 UTC 2010


On 06Jul2010 09:47, Felipe Contreras <felipe.contreras at gmail.com> wrote:
| On Tue, Jul 6, 2010 at 2:18 AM, Cameron Simpson <cs at zip.com.au> wrote:
| > On 05Jul2010 20:02, Felipe Contreras <felipe.contreras at gmail.com> wrote:
| > | I mean that there are many steps involved, go to this page, fill this,
| > | wait for that, reply here, etc.
| >
| > The procedure of: fill in the joining form, reply to a confirmation
| > email is a standard step in most mailing lists these days.
| 
| But the actual steps you have to do are different depending on the
| list manager; majordomo is different than mailman.

So? It is not _very_ different.

| > It serves two
| > purposes: it catches misspelt email addresses (because the confirmation
| > email fails) and it rejects most spam robots (because they don't process
| > the confirmation email, and generally don't provide anything but lies in
| > email address fields anyway).
| >
| > If there were no spammers and no misconfigured user mail clients and no
| > misfilled forms then this step would not be necessary. But it is because
| > is so greatly reduces trouble.
| 
| Yes, but again; other mailing lists manage just fine without requiring
| subscription.

Good for them.

| > | >> Public mailing lists should receive mail from *anybody*; if the poster
| > | >> is not subscribed, then the message should go through moderation. This
| > | >> is the truly open way.
| > | >
| > | > No thanks.  If you want groups full of spam, there's usenet for that.
| > |
| > | There's filters for that. If your current filter doesn't work, switch.
| > | bogofilter maybe?
| >
| > Gah! NO.
| >
| > This is the standard answer of the spammers too. It is the difference
| > between "opt-in" and "opt-out" junk mail. Advertisers always argue that
| > they should be free to put people on their mailing lists and that people
| > should opt-out when they get the junk. This places the burden on every
| > individual receiver of the mail, and is unreasonable.
| 
| No, the spam filter will be *in addition* to whatever is there
| already.

So, server end? Good. But spam still needs to be winnowed from
probably-spam and from "unknown poster" (yeah, any list I run counts
those as of unknown value, and maybe spam).

| If the list remains subscription-only, there's still spam
| that goes through, the spam filter will help. And if the lists is
| moderated, the spam filter would help go through the moderation queue.
| 
| The burden would not be on the subscribers, in fact they would receive
| less spam.

Ok, but the burden on the moderators goes up. Putentially WAY up.

| > | > Subscriptions is a step in minimising crap being posted to the list
| > | > (whether that be spam, or simply tossers who'll post rubbish to lists,
| > | > just to spout crap from their fingers).
| > |
| > | Really? So I don't subscribe I'm a looser whose posts are not welcome?
| >
| > Well, if you don't subscribe you're too lazy to meet the very low bar to
| > entry to the discussion; maybe you're not desirable. This isn't so in
| > your case, since you're clearing prepared to argue cogently for your
| > point of view. But for the many many spammers it _is_ the case that they
| > are not welcome.
| 
| Again, that's speculation.

No, it's not. It is empirical fact. All the "maybes" are speculation on
an individual post basis, but in the aggregate the distribution is real.

| Most of the mailing lists I'm subscribed
| to, allow non-subscribers to post, and there are as many occasional
| posters who don't have a clue there, than here. So again, I don't
| think it's sensible to apply prejudices based on the people's
| subscription, which is very simple to do.

Nobody is telling you how to run _your_ lists.

| > | And if some random guy manages to subscribe (which according to a
| > | previous post it's easy), then it's post is worthwhile?
| >
| > It means they've made an effort: they are probably not a spammer and
| > probably not a robot and _are_ probably more motivated to participate in
| > a valuable way. So yes, _probably_ they are more worthwhile.
| 
| The correlation between membership and worthwhileness is very week at
| best. Again, the person can subscribe, troll, and leave.

Of course. But the correlation is not that weak. At least only people
prepared to make a (small) effort get to post. Spammers make ZERO
effort.

[...]
| > | > If the list was moderated in the way you propose, moderators would spend
| > | > all their spare time checking new messages, and it'd be ages before your
| > | > post got through.
| > |
| > | No, it takes time, but eventually it gets posted. I do this on many lists.
| >
| > Not on all lists. Moderators often lack the time. I speak as one such,
| > and some posts simply don't make it because they have waited too long
| > for attention. It is not scalable. If the list is very active, the
| > problem gets far worse.
| 
| Again, other lists manage just fine. Speculation.

Sigh. Not speculation. I have seen lists made useless by spammers, and
abandoned, and lists where the burden becomes too much for the
moderators. It does happen.

| > | >> Orthogonal to this is that the mailing lists should not mingle with
| > | >> "Reply-To"; they should leave the To and Cc fields intact, so that the
| > | >> MUA can reply to the right addresses. See:
| > | >> http://www.unicom.com/pw/reply-to-harmful.html
| > | >
| > | > The list works fine.  Messages go to the list, and the list's address is
| > | > in the "to" field, where it belongs.
| > |
| > | The initial mail might have the ml is "To", but might also be in "Cc".
| >
| > Shrug. For mail clients that distinct is usually irrelevant.
| 
| No. Imagine you have a discussion with a group of people, most of them
| are not subscribed to the ml. Then you think; hey, why not Cc the
| fedora-user mailing list and see what they think?

You think: "why not _spam_ the fedora list with some discussion whose
ancestors posts have been seen by nobody".

| Well, the Reply-To
| munging will override the Cc and make all the replies go to the ml
| *only*.

Your mail client is weak.

| Sure, MUAs don't make much of a difference between To and Cc (people
| do), but Reply-To makes a huge difference.

Can do, for the plain "reply" setting. That is not your only choice.

| > | But the later posts wouldn't if the rules were right. I would be
| > | sending this mail to you, as you were the originator, and the mailing
| > | list would be kept in Cc.
| >
| > Plenty of people don't want a personal CC; they want all replies on the
| > list only.
| 
| Really? Why? Are you speculating again? Where's the evidence?

Gah. I haven't been speculating anywhere. _All_ the stuff I've said is
based on things I have personally seen.

Including plenty of people who explicitly ask not to be personally CCed,
yea, even to the point of putting such a request at the top of _every_
post they make.

| I have never seen a single person complain about it.

My experience is broader than yours then.

| > | This has another advantage. Many clients, like Gmail, do smart thing
| > | where the user is in the To field (or Cc for that matter),
| >
| > It might help if you said what you think _is_ the smart thing. Since
| > the current setup of this list _is_ what is often wanted, some
| > justification is needed _after_ you describe "the smart thing".
| 
| Gmail shows personal level indicators:
| http://mail.google.com/support/bin/answer.py?hl=en&answer=8156
| 
| Also, you are just assuming that whatever setup this mailing list has
| is what is often wanted. Another possibility is that people really
| didn't think too much about it.

Sure, that was a possibiity. But based on the reponses in this thread,
it is demonstrably false. It is _not_ wanted.

| > | but on
| > | Fedora lists, that information is lost thanks to the Reply-To munging.
| >
| > Bah. Nothing need to use it.
| 
| Maybe not you, but other people want to know when the mail was sent to
| the mailing list, or both to the mailing list and them.
| 
| > | In my MUA, notmuch, for example, I have a rule; if the message was
| > | sent to the the certain mailing list, skip the inbox tag, unless it
| > | was sent to me. No way to do that here.
| >
| > Demonstrably false. I file my email in _exactly_ the same way, and it
| > works just fine.
| 
| Impossible. You can't filter mail that was addressed to you and the
| mailing list. If the Reply-To header is munged, the mail will appear
| to be addressed to the ml only.

Funny, my filters work. I'm beginning to think you live in a fantasy
world.

| > I _suspect_ you mean that you want the post sent to you directly in
| > addition to the list (though you have not said it). Presumably so you
| > can track the discussion more easily, and that is a rasonable desire
| > (the tracking ability, not the particular implementation).
| 
| That is exactly what happens when the Reply-To headers are not munged!
| 
| > There are other ways.
| >
| > For example, I colour threads in which I have participated distinctively; I
| > can then see them easily when browsing my "unix lists" mail folder.
| >
| > Also, I _always_ press group-reply/reply-to-all and never plain reply,
| > and then I consider the resulting to/cc headers and edit if suitable.
| > That bypasses your reply-to issue altogether.
| 
| Sure, for *outgoing* mail I can override the Reply-To filed. But when
| people hit reply, the reply will go to Reply-To (the ml only), and
| then I would not be able to tell if the mail was addressed to me or
| not.

Sure you can. It wasn't addressed to you. Easy. In fact, that's what
you're complaining about I think.

But you could in principle track message ancestry through the In-Reply-To
header.

| > | > Because that's where the post was
| > | > sent to.
| > |
| > | Only because of Relpy-To munging; it's a vicious circle.
| >
| > And yet often desirable.
| >
| > Plenty of people are incapable of keep the discussion on list (with
| > or without cc-to-author). Yea, even my mother in personal email often
| > drops the CC to my partner. The reply-to makes that work for this list.
| 
| Speculation. Other mailing lists manage just fine.

Again, NOT speculation. I have seen this in the real world; I'm even
citing specific examples for you. You seem to just ignore stuff that
doesn't fit your world view.

| If your mother drops the Cc, that's your mother's fault. But on public
| open source mailing lists, people know the difference between reply
| and reply to all, or they will learn, just like they learn netiquette.

Many never learn. And if they can learn that, they can learn to cope in
the face of lists that munge the repy-to.

| > | > Lists that don't put themselves in the reply-to end up with very few
| > | > posts coming back to the list.  You see a list full of the same
| > | > questions being asked over and over, because there's no replies being
| > | > made in public.
| > |
| > | You are speculating, and doing it wrong.
| >
| > He's not speculating, it is empirically true. The degree to which it is
| > true varies, but the effect is real and not small.
| >
| > | Most of the mailing lists I'm subscribed to don't munge the Reply-To
| > | header, and all the threads are kept in the ml (in fact multiple
| > | mailing lists as cross-posting works).
| >
| > You can't tell how much escapes, because you do not see the escapes.
| 
| Well, you can't tell either, can you? It might might as well be 0. If
| you say it's not 0, that's speculation as you have no proof.
| 
| > So it may well largely work, but it is unreliable.
| 
| If it was a problem, somebody would have complained, don't you think?

Sure they complain, and when they set up lists of their own they minge
the reply-to to avoid the issue in future.

| > | > You can still reply to the right address.  A default reply will come
| > | > back to the list, where it's supposed to (for this list).  You can opt
| > | > to reply to the person by replying to their *from* address, because the
| > | > poster's address is in the place that it ought to be, the "from"
| > | > address, because that's where the message came from.
| > |
| > | MUA's don't do that. Reply-To overrides everything. If Reply-To wasn't
| > | overridden, there would be the option to reply to the person, or reply
| > | to everyone (including the ml).
| >
| > Demonstrably false. My MUA has those choices. And I use the second one and
| > edit where it turns out to be not desirable (infrequent).
| 
| What are you talking about? I said that the MUA would present two
| choices, and you are saying that you have the two choices... So what I
| said is true.

Go and reread what you said above. You didn't say that.
If you meant to say it, we may agree.

| > | > I can't recall whether it changes the CCs, and I half agree with keeping
| > | > them.  Unfortunately some troublemakers abuse that, by replying to some
| > | > post, and adding inappropriate addresses to a CC field.
| > |
| > | Reply-To overrides the Cc too.
| >
| > _If_ you use the mailer's "conventional reply" operation. And even then,
| > the outcome is more right than not.
| 
| When Reply-To is munged, you don't have the "reply to all" option,
| only "reply". And I said, the Cc is overridden.

Only if your MUA is weak. You can reply to all the people in the to/cc
any time you like.

Maybe your complaint is that the to/cc is just the list; you didn't make
that clear either.
-- 
Cameron Simpson <cs at zip.com.au> DoD#743
http://www.cskk.ezoshosting.com/cs/

Facts do not discourage the conspiracy-minded.
        - Robert Crawford <rawford at iac.net>


More information about the users mailing list