[Mailman-Users] Allowing users to join without specifying pas swords

[Chuq Von Rospach]

On Friday, June 15, 2001, at 11:12 AM, Barry A. Warsaw wrote:

> That's what I wanted to know!  If I understand one thing, its that my
> own demands on Mailman are heavily skewed to the atypical lists I run,
> the systems I run it on, and the kinds of tasks it forces on me.  So
> getting other points of view are critical here (which I why I love
> this community!).

we've made the decision that all of our mail systems are going to be 
sent individually. I'm planning on putting the subscribed address back 
in the To: line, encode for one-click unsub, and encode data to make it 
easy to figure out what the subscribed address is elsewhere in the 
message. Users are increasingly unclear about the underlying aspects of 
mail lists and BCC, and between naive users, people who leave a trail of 
.forwards across the universe and forget about them, and mail servers 
that do stupid things with email (not that I'm calling out First Class, 
Notes or Exchange as braindamaged, but if the shoe fits...), it'll make 
everyone's life better.

Besides, in all honesty (and this isn't necessarily a factor with 
Mailman, as opposed to my big server) we want to distance from the bulk 
feeling of email more towards personalization and 'first class' letter. 
But I think the underlying issues are still applicable to mail lists.

Whether it's naive users, users with 37 hotmail addresses, the person 
who's MIT alum account forwards to his UCLA grad student account that 
forwards to his bigfoot account that forwards to his hotmail address 
that forwards to his DSL line, it's simply not safe to assume that (a) 
the address the person is mailing from is the address subscribed, (b) 
the user will actually know what address is subscribed, or (c) that the 
user is interesting/able to figure this out with you.

And that ignores large companies who's IS departments re-arrange email 
addresses on a regular basis, and don't bother to tell their users. I 
can't tell you how often I get mail from fred.luser at big-company.com, 
who's actually subscribed as fred-luser at mail.big-company.com, and he 
didn't even know about the change until I mentioned it to him...

So from the point of view of protecting the list server from the rest of 
the universe and making the list-admin (and site admin)'s jobs easier, 
lists need to have the subscribed address in them and accessible, 
because there are simply too many opportunities now that create a piece 
of email that got to a user, and there simply isn't any way of finding 
out how. And you usually don't find out until after the user is truly 
and legitimately frustrated at you and your server anyway, which is a 
great way to START a discussion....

> So what should the default be?  While I appreciate this sentiment (a
> lot!), I'm still worried that if we turn this on, a naive sys admin
> could get themselves in hot water, and start flooding us with "Why
> does my performance suck so bad?!" messages.

if you go check your archives, you'll see we already get those, but it's 
because mailman sets that number large (500, I think), and therefore 
slows delivery on systems where the MTA intelligently breaks it up for 
you (which postfix can, but sendmail won't).

I'd set this number very low, like 5 for 2.0.x, and 1 (effectively VERP) 
for 2.1 -- and then document it in the code and FAQ that it should be 
raised only if the server load isn't sustainable (but in 2.1, here are 
the features you lose....), and only the smallest amount needed to get 
it under control until you add capacity... I think you have many fewer 
problems doing that than setting it large and expecting people to tune 
it down.

The real issue here, IMHO, is likely ot be the wonderful sendmail/DNS 
lookup on acceptance delay issue. And I'd suggest the way to fix that is 
to fix the Sendmail module, and have it use -Odeliverymode=defer. and 
use the Smtp module for other MTAs. If they want to remote-SMTP to a 
sendmail MTA, document that they're screwed. That's the only way to get 
around this sendmail chokepoint, other than to do what I'm doing, and 
move to postifx...


--
Chuq Von Rospach, Internet Gnome <http://www.chuqui.com>
[<chuqui at plaidworks.com> = <me at chuqui.com> = <chuq at apple.com>]
Yes, yes, I've finally finished my home page. Lucky you.

Any connection between your reality and mine
is purely coincidental.