clang-3.6 build failure

[Clemens Lang]

Hi,

----- On 22 Dec, 2014, at 14:16, René J.V. Bertin rjvbertin at gmail.com wrote:

> Most of my other lists make a reply go to the list, because that's what you
> usually want (as also evidenced by the couple of "keep this on the list" I've
> gotten)

I still don't like this behavior. It might be easier for newcomers or people not
used to mailing lists, but it actively breaks my MUA's behavior.

>> "Reply to List" goes to
>> the list if any,
>           ^^^^^^
> 
> Delivered-To: rjvbertin at gmail.com

The problem is right there – GMail deduplicates mails and only ever keeps a
single copy per Message-ID. You would, however, have received multiple copies
of this mail, one via the explicit To/Cc, and one via the list. The copy that
went through the list would have had the header, the copy that went to your
address directly does not. There is also a mailman setting that avoids sending
you an email again if it already sees your address in To or Cc.

To avoid this, you could, for example, set a Reply-To header to point to the
list to indicate that you don't want to be addressed directly in a reply.


> That aside, I don't really care how it's done, but I won't be
> blamed if I forget to do a Reply-all and then remove the superfluous addressees
> when sending a reply that isn't itself something I re-read twice to be sure I'm
> not submitting nonsense.

That's fine. Let the receivers deal with superfluous copies of mails they receive
due to them being listed in To/Cc.

>> "Reply All" goes to everyone. Faking headers to change
> 
> Adding a header that's not already present isn't faking, it's setting a
> (hopefully sensible) default.

Not necessarily. The absence of the header is valuable information in itself. For
example, if I add the Reply-To header to the list, I'm explicitly requesting any
answers not be sent (additionally) to my address, but to the list only. Adding
this header for somebody implies that's what the poster wanted, but that may not
necessarily be the case.

Also note that munging headers becomes increasingly dangerous now that DKIM
signing is getting more widespread. Most configurations I've seen don't sign the
Reply-To header, but it would be perfectly valid to do that as well, which would
mark any copies re-distributed via our list with a mangled reply-to header as
spam.

-- 
Clemens Lang