"D. J. Bernstein" <[EMAIL PROTECTED]> writes:
| Scott Schwartz writes:
| > I think it's strange that qmail-inject uses '-' to seperate the mailbox
| > from the verp part, even when some other conf-break character is in
| > effect for that user. This surely violates the principle of least
| > surprise,
|
| Certainly not. If a mailing list is named
|
| Jim+Joe-Bob#3
|
| then people expect the service addresses for that list to be
|
| Jim+Joe-Bob#3-request
| Jim+Joe-Bob#3-owner
| Jim+Joe-Bob#3-subscribe
|
| and so on. What follows Jim+Joe-Bob#3 is _always_ a dash. It doesn't
| matter whether these addresses are controlled by Jim, or Jim+Joe, or
| Jim+Joe-Bob, or Fred.
Most people traditionally think of list-request as a seperate alias,
not as a submailbox, because user names can and do have hyphens in
them, and because mailing lists are usually not users. You want to
conflate those usages, though, and so you are arguing here that
conf-break should always be '-'. However, you're also the one who
decided that conf-break is allowed to be something else, and lots of
people do set it to something else.
The problem is that if conf-break isn't '-', then qmail-inject's "m"
and "r" options will turn replyable addresses into VERPs that are not
replyable addresses (ironic, considering VERP's ostensible purpose).
That's pretty damn surprising.
It's (doubly) surprising because qmail makes a big deal about
user-controlled subaddresses, and the manpages say what conf-break is
set to and talk about how users/assign works, and so people who ask
qmail-inject to generate a VERP probably expect it to generated a
usable one without having to also manually insert subaddresses into the
envelopes of outgoing messages.
Moreover, a VERP isn't intended as a rendezvous, entirely unlike
list-request addresses. Outsiders have no idea what conf-break is on
some other system, nor should they. On the other hand, the person who
generates a VERP does know what conf-break is, so the surprise is if
the VERP generator ignores conf-break, not if conf-break differs from
some convention in a documented way.
If you disagree, then, as I suggested in private mail, I think you
ought to augment the documentation to warn people that changing
conf-break will cause them extra hassles.
