Re: various nnml questions, many about cross-posted mail messages

[jbw@cs.bu.edu (Joe Wells)]

In article <x64tolr58d.fsf@eyesore.no> Lars Magne Ingebrigtsen <larsi@ifi.uio.no> writes:

   jbw@cs.bu.edu (Joe Wells) writes:

   > 1. An article is cross-posted to two mail groups:
   >    
   >      Xref: csb.bu.edu mail.emacs.bbdb:1 mail.to-jbw:3
   > 
   >    I am in the Summary buffer for mail.emacs.bbdb on the line for this
   >    article.  I type "B DEL".  The file ~/Mail/mail/emacs/bbdb/1 is
   >    unlinked.  The file ~/Mail/mail/to-jbw/3 is untouched.  In particular,
   >    it still has the old "Xref:" header, which is incorrect because the
   >    article is no longer in mail.emacs.bbdb.
   > 
   >    Questions:
   > 
   >      A: Is this fixed?

   No, but it isn't a bug.  :-)

Hmm.  All it leaves is a dangling pointer, which is okay since newsreaders
have to be prepared to handle this.  So I agree here.

However, I think it would be better if Gnus fixed the Xref header in the
other copies of the message.

   >      B: Is there a command that will remove the article from all of the
   >         mail groups to which it is cross-posted?

   Hm.  Nope, but it's trivial to add.  I've now added it to the Red Gnus
   todo list.

   > 3. An article is cross-posted to two mail groups:
   > 
   >      Xref: csb.bu.edu mail.emacs.bbdb:2 mail.list.info-bbdb:1
   > 
   >    I am in the Summary buffer for mail.emacs.bbdb on the line for this
   >    article.  I type "B m" and specify the group "mail.misc".  The file
   >    ~/Mail/mail/emacs/bbdb/2 is unlinked.  The file ~/Mail/list/info-bbdb/1
   >    is untouched.  In particular, it still has the old "Xref:" header,
   >    which is incorrect because the other copy for the article is no longer
   >    in mail.emacs.bbdb, but is instead in mail.misc.

   Well, it's not incorrect -- it just points to a message that no longer
   exists.

In this case, I disagree with you.  What does it mean when a user types "B
m" in a mail group?  It means they wish to change the result of their
nnmail-split-methods method.  Thus, if a message was put in mail groups A,
B, and C by nnmail-split-methods and the user types "B m D RET" in group
B, then the result should be as though nnmail-split methods put the
message in groups A, D, and C instead.

I also believe that a newsreader, to the greatest extent practically
possible, should treat a message cross posted to several groups as a
single entity.  When I receive a message that gets put in groups A, B, and
C, I do not want it to make a difference in which group I deal with the
message first.  If, while in group B, I read the message, tick it, and
reply to it, I expect it to be annotated with "!A".  If a week later I am
viewing the same message in group C, I want to see the same annotation
there too.

In order for a newsreader to meet the goals I mentioned in the previous
paragraph, it is necessary for that newsreader to do as much as it can to
preserve the semantic integrity of the Xref links.

   >    The new file in
   >    mail.misc has the following wrong "Xref:" header:
   > 
   >      Xref: csb.bu.edu mail.misc:48
   > 
   >    This is wrong because the message is cross-posted to
   >    "mail.list.info-bbdb".
   > 
   >    Questions:
   > 
   >      A: Is this fixed?
   >      B: Is there a command that will take an article which is cross-posted
   >         to groups A and B and make it cross-posted to A, B, and C?  The "B
   >         c" command does not seem to preserve cross-posting links.

   Should it?  I don't know, really.  Again, changing the behavior is
   easy enough, but I'm not sure it's the right thing to do...

See above.

   > 5. An article is cross-posted to multiple mail groups:
   > 
   >      Xref: csb.bu.edu mail.emacs.bbdb:4 mail.emacs.mail-extr:1 mail.to-jbw:22
   > 
   >    I am in the Summary buffer for mail.emacs.bbdb on the line for this
   >    article.  I type "!" to tick the article.  I exit this summary and
   >    enter the Summary buffer for mail.emacs.mail-extr.  There is no tick
   >    mark for this article in the group mail.emacs.mail-extr.
   > 
   >    Question:
   > 
   >      * Is there a variable setting so that any mark change (adding or
   >        removing a mark) I apply to a message in one group is immediately
   >        applied to the same message in any other groups to which it is
   >        cross-posted?  Right now only the "R" marks seem to be propagated.

   This is a feature.  Propagating marks between groups seems rather
   absurd to me.  I can't see what good that would do.  

See above.

   The only thing the Xref utility is there for is to avoid having to
   read any article more than once, and that's all.

First, I disagree about the usefulness of Xref.  Second, the lack of mark
propagation in Gnus prevents achieving the goal of not reading articles
more than once.  Suppose, while in group A, I read an article which was
cross posted to group B and then I tick the article.  Then I go to group
B.  I will see the same article and it will be unmarked (not marked as
read).  Unless I am very careful, I will end up reading it again.

   > 2. In the Summary buffer, the "B m" command asks for a target group.  I
   >    type in a name.  It complains that the group doesn't exist.  (The group
   >    does indeed not exist yet.  I don't care.  I want the message in that
   >    group.)
   > 
   >    Question:
   > 
   >      * Is there a command that will just create the group instead of
   >        failing?

   No, the group has to exist before moving.  I could remove the
   restriction; I just thought that requiring a match would lead to fewer
   problems with typos and such.  Whadda y'all say?  Should Gnus just
   create the group when the user moves/copies an article to the group?

For me, typos are not a problem because I usually use completion to verify
that what I am typing is correct.  Even when I type all of the letters of
a string by hand, I still use TAB or ? to check that it is right.  So if I
type "mail.foo.bar", I really mean "mail.foo.bar".  For me, it can not
possibly be a typo.  Other people might not share this style.

   > 4. Question: How do people keep track of whether they have edited a
   >    message or forwarded it?

   They have rilly good memories?

I asked this question because this is one of the features of VM which I
like quite a bit.

-- 
Joe Wells <jbw@cs.bu.edu>