Re: [Nmh-workers] Format function to create wrapped header lines?

[Robert Elz]

    Date:        Mon, 05 Sep 2016 21:18:38 -0400
    From:        Ken Hornstein <address@hidden>
    Message-ID:  <address@hidden>

  | You had asked if anyone actually uses the References header for threading,
  | and it turns out the answer here is "yes!".

OK, though it really ought be used only to help find messages that
might be part of the thread, not to define it.

  | >Some of the issue here, is that there's no concept of a "thread" in the
  | >e-mail standards, and everyone defines it slightly differently.
  | 
  | Yeah, I mean ... I think that came over from Usenet, right?

As I recall, yes.

  | I guess it's a set of messages which are all replies to one another.

In that case, just using References is incorrect (trn certainly never did 
that!) as that's not what references is.   Neither my previous message, nor
this one from you I am replying to were (in any sense) part of a set of
messages which included Ralph's 2015 message as being replies to one other.
Yet my previous message (correctly) and then your reply (arguably incorrectly)
included a References: to that message.   (I've deleted again this time, though
arguably in this message it should be retained.)

References should be like that section at the end of a paper in a conference
or journal, where you list the identities of all the earlier papers that your
paper refers to (substituting message for paper.)  That cannot simply be all
of the References from each earlier paper (message) with a reference to the
most recent one tacked on...

But, as I said last time, doing it correctly is HARD, and no-one really
usually wants to bother.   However, that doesn't mean that anyone should
start assuming that every message in a thread will be in the References
header of the most recent message - some of the messages in References will
be earlier messages in the thread, others won't be part of the thread at all
(sometimes) and there can be messages that are part of (even the linear)
thread, that are not included in the References.

Using the References header is just fine - just don't start making assumptions
about what it will (or should) contain that conflict with how the header
should really be being used.

kre