8965 - [FEATURE] No support for In-Reply-To header

Status: VERIFIED FIXED

(MailNews Core :: Backend, enhancement, P3)

Description

[tenthumbs]

It's an RFC822 header and lots of MUAs use it. Some, like Pine, use it
exclusively when replying to a message. Since Mozilla doesn't support it, any
reply to a message from a Pine user prevents Mozilla from properly threading.

I think failing to support this header is a bug.

Comment 1

[Phil Peterson]

Actually, it's an enhancement request, but we track those in bugzilla too.
Reassigning to bienvenu since he's The Man for threading.

David, there's a thread on In-Reply-To usage conventions here:
news://news.mozilla.org/375D0CEC.43C710C3%40cybernex.net

Comment 2

[David :Bienvenu]

Well, we'll see what we can do.

Comment 3

[tenthumbs]

There's an even earlier thread starting at
news://news.mozilla.org/[e-mail] .

If you saw the mess it made of the linux-kernel mailing list, you'd agree it's a
bug.

Comment 4

[lchiang]

correcting platform to All since if addressed, this will affect all platforms.

Comment 5

[David :Bienvenu]

I don't think we can do this for news, since the in-reply-to header is not part
of xover data, but we can try to do it for POP3 and IMAP.

Comment 6

[tenthumbs]

It's an RFC822 header so it's only for mail. It's also optional. I'm not
suggesting emitting one only reading and using it if it's there.

Comment 7

[David :Bienvenu]

moving to m16

Comment 8

[David :Bienvenu]

adding feature to summary. We hae to do this.

Comment 9

[David :Bienvenu]

fix checked in. you will need to delete the .msf file if you want existing 
messages with in-reply-to headers to get rethreaded.

Comment 10

[tenthumbs]

I'm fairly sure that In-Reply-To and References are not mutually exclusive in
email so using one or the other but not both may not work. I'll gather some data
later.

Also, you are prepared to handle garbage in either header, right?

Comment 11

[David :Bienvenu]

It's fine if an email uses both; I'll just use the REFERENCES header. If an 
email generates an incorrect REFERENCES header (i.e., w/o the message-id of the 
replied-to message) that seems like a bug in the original email client.

I'm neither more or less able to handle garbage in either header than I was 
before. We don't crash if we can't find the message id, if that's what you mean. 
But neither do I fall back on the in-reply-to header if the references header is 
invalid.

In the messages I looked at (probably from Pine), there was an In-Reply-To 
header and no references header. Which is what caused all the problems in the 
first place, if I understand correctly.

Comment 12

[tenthumbs]

RFC822 allows non-message-id text in both the In-Reply-To and References
headers. It actually happens in practice. For best results, it is necessary in
email (not news, of course) to extract message ids from the headers. If you are
already doing that, then there should be no problem. Pine turns out to be rather
well-behaved in the sense that it only includes message ids in its i-r-t header.
Other MUAs aren't.

I have some data from my mailing lists which shows all sorts of variety. It's
available if you want it.

Comment 13

[lchiang]

tenthumbs - fixed for you?

Comment 14

[Ben Bucksch (:BenB)]

I still get weird threading with with In-Reply-To headers, which is related to
giving |References| the preference, IIRC. Will investigate.

Comment 15

[David :Bienvenu]

yes, we've discussed that on the newsgroup. I doubt I'm going to have any time 
to do more than what I've done. If anyone wants, they can open a different bug 
for giving the references header preference and not throwing out invalid 
references.

Comment 16

[tenthumbs]

Certainly greatly improved but I see weird threading also. I found at least one
mailer that does this:

X-Mailer: exmh version 2.1.1 (devel)
In-Reply-To: <[e-mail]>
References: <[e-mail]>
  <[e-mail]>
  <[e-mail]>

The message content indicates that the message should thread against the
In-Reply-To header but Mozilla threads against the last id in References.

Has another bug been filed on this?

Comment 17

[David :Bienvenu]

no, that was a choice I made to prefer the references header.  I've got to
choose one to prefer, and if I chose to prefer the in-reply-to-header, I'm sure
someone would come up with a case where I should have prefered the references
header.

Comment 18

[tenthumbs]

You're absolutely right that there's no way to win. The only question is how
many people will complain. Maybe this should be a pref.

Comment 19

[David :Bienvenu]

the problem with making it a pref is that you could get mail from clients that 
are broken in either way, so the pref would just be saying which brokeness you 
want to favor, which seems of limited value.

Comment 20

[tenthumbs]

Sure, but at least you could find the setting that works best for you. Empower
the user and all that stuff.

Comment 21

[esther]

This bug was fixed for original scenario in April of 2000.  The reporter
commented in June of 2000 that this works better.  There was some discussion as
to what else should be fixed and was suggested new bugs be logged for specific
outstanding issues.   Continued discussion about what is and what should be
ended June 8 2000 without any resolve.  I will verify this as fixed per the 1st
fix.  If the reporter or any others commenting logged new bugs for specific
outstanding issues, please note them in this bug.