Archived sent messages

Archived sent messages

[Lars Magne Ingebrigtsen]

I had a look over the methods that Gnus use to save outgoing messages,
and I didn't much like what I saw.  `gnus-mail-self-blind',
`gnus-author-copy', "Fcc", and `gnus-outgoing-message-group'.  They
all seem so, like, unintegrated.  I want mail I've sent to appear as a
group automagically, wihtout me having to fiddle with all these
variables and then having to do a `G m' to create a group manually.

So I've added a few variables and a few functions.

`gnus-message-archive-method' is a virtual server that stores all
messages you've sent.  It is native in a way -- it'll always be
queried for new groups and new articles in the groups, so no fidding
with it is necessary.

`gnus-message-archive-group' is a variable that says what group on the
archive server to say in.  It can be a string, or a list of strings,
or an alist of regexps, functions, and forms.  (The usual.)  It is 

  '((if (eq major-mode 'news-reply-mode) "misc-news" "misc-mail))

by default, which means that articles are saved in "misc-news" and
mail is saved in "misc-mail", which seems reasonable to me.

All this is on by default, which means that new users will get
archiving by default.  The rest of us who use other archiving methods
either have to turn this off, or turn the older archiving methods
off.  Or get double archiving. 

-- 
   Lars Magne Ingebrigtsen * larsi@ifi.uio.no
      (a red leaf that falls from the purple tree)

Re: Archived sent messages

[Per Abrahamsen]

>>>>> "LMI" == Lars Magne Ingebrigtsen <larsi@ifi.uio.no> writes:

LMI> It is 

LMI>   '((if (eq major-mode 'news-reply-mode) "misc-news" "misc-mail))

LMI> by default, which means that articles are saved in "misc-news" and
LMI> mail is saved in "misc-mail", which seems reasonable to me.

I think the default should be to let `nnmail-split-fancy' decide which
group the outgoing messages go to.  Having in and outgoing messages in
the same groups is really nice in a threaded mail reader, and I
believe it would be a good default for new users.

It would also be very nice if the outgoing messages were automatically
marked as read.

Re: Archived sent messages

[Steven L. Baur]

>>>>> "Per" == Per Abrahamsen <abraham@dina.kvl.dk> writes:

>>>>> "LMI" == Lars Magne Ingebrigtsen <larsi@ifi.uio.no> writes:

LMI> It is 

LMI> '((if (eq major-mode 'news-reply-mode) "misc-news" "misc-mail))

LMI> by default, which means that articles are saved in "misc-news" and
LMI> mail is saved in "misc-mail", which seems reasonable to me.

Per> I think the default should be to let `nnmail-split-fancy' decide which
Per> group the outgoing messages go to.  Having in and outgoing messages in
Per> the same groups is really nice in a threaded mail reader, and I
Per> believe it would be a good default for new users.

Problem: Mailing lists which return messages to the sender.  You will
find yourself with duplicate message IDs in your nnml group which Gnus
doesn't handle very well at present.

I'd like to see retain an author copy on such lists only so long as
the message hasn't come back from the list server, then the author
copy can be replaced with the copy from the list.

Per> It would also be very nice if the outgoing messages were automatically
Per> marked as read.

Yes and no.  Getting at read messages is a FAQ.  Perhaps this should
be a variable, as I could see a good argument for making the articles
dormant too.

-- 
steve@miranova.com baur

Re: Archived sent messages

[Per Abrahamsen]

>>>>> "LB" == "Steven L Baur" <steve@miranova.com> writes:

LB> Problem: Mailing lists which return messages to the sender.  You will
LB> find yourself with duplicate message IDs in your nnml group which Gnus
LB> doesn't handle very well at present.

LB> I'd like to see retain an author copy on such lists only so long as
LB> the message hasn't come back from the list server, then the author
LB> copy can be replaced with the copy from the list.

I think the following variable is supposed to do that.  

I'd feel more comfortable with it if it stated that it would do some
sanity check to ensure tha the articles really are the same, and that
the local copy rather than the remote copy was deleted.

nnmail-delete-duplicates
  Variable: *If non-nil, nnmail will delete any duplicate mails it sees.
  Plist: 1 property

>>>>> "Per" == Per Abrahamsen <abraham@dina.kvl.dk> writes:

Per> It would also be very nice if the outgoing messages were automatically
Per> marked as read.

LB> Perhaps this should
LB> be a variable, as I could see a good argument for making the articles
LB> dormant too.

Yes, dormant is even better than read.

Re: Archived sent messages

[Lars Magne Ingebrigtsen]

Per Abrahamsen <abraham@dina.kvl.dk> writes:

> I'd feel more comfortable with it if it stated that it would do some
> sanity check to ensure tha the articles really are the same, and that
> the local copy rather than the remote copy was deleted.
> 
> nnmail-delete-duplicates

I've renamed that variable and revised the duplication handling
scheme.  The variable is now called `nnmail-treat-duplicates' and is
`warn' by default.  When nnmail finds a duplicate message (by keeping
a cache of Message-IDs), it will insert a `Gnus-Warning' header and
rename the Message-ID header.  This means a) that duplicates will now
start to appear in the summary buffers, and b) they will be labeled as
such.

You can set this variable to nil to get the old behavior (don't check,
don't show); and `delete' to remove duplicated messages.  

No sanity checks are performed; the Message-IDs are just compared.

-- 
Home is where the cat is.

Re: Archived sent messages

[Per Abrahamsen]

Maybe it should be possible to set `nnmail-treat-duplicates' to a
function that a list of pointers to the messages as args, and returned
a list of actions to be performed on these messages (warn, delete,
ignore).  Then we could experiment with adding sanity checks and so.

Re: Archived sent messages

[Lars Magne Ingebrigtsen]

Per Abrahamsen <abraham@dina.kvl.dk> writes:

> Maybe it should be possible to set `nnmail-treat-duplicates' to a
> function that a list of pointers to the messages as args, and returned
> a list of actions to be performed on these messages (warn, delete,
> ignore).  Then we could experiment with adding sanity checks and so.

In September 0.28 that variable can be a function.  The function will
be called from a buffer narrowed to the message in question with the
Message-ID as a parameter in nnmail suspects that the message is a
duplicate.  It should return one of `warn', `delete' or nil.

-- 
Home is where the cat is.

Re: Archived sent messages

[Brian Edmonds]

>>>>> "Steven" == Steven L. Baur <steve@miranova.com> writes:

Steven> Problem: Mailing lists which return messages to the sender.  You
Steven> will find yourself with duplicate message IDs in your nnml group
Steven> which Gnus doesn't handle very well at present.

I'm afraid I don't see the problem.  With gnus-bbdb you just define a
public and private folder for each list address.  By checking the
envelope sender it can then file mail to the list in one mail group, and
mail from the list in another.  I have trouble imagining maintaining a
really good regexp set for mail filing without gnus-bbdb.

That way I keep an archival copy of whatever I send (in bwe.emacs.misc
for the ding list, for example) and get a copy back into the usual list
group which is eventually expired (bwe.emacs.lists.ding for this list,
for example).  The same thing happens with private copies from other
people that are also sent to lists, though they file based on who it's
from.  Preto kerplong, no duplicates in long term storage.

I'm not sure if the best approach for newbies is to have Lars' outgoing
mail archive be the default, or to let the default be no outgoing
archiving, as one would normally expect.  I lean towards the latter.
Having Lars' automated approach available as an option is certainly
nice.

Steven> I'd like to see retain an author copy on such lists only so long
Steven> as the message hasn't come back from the list server, then the
Steven> author copy can be replaced with the copy from the list.

Yeah, right.  Good luck determining this in any automated way.

Brian.

Re: Archived sent messages

[Steven L Baur]

>>>>> "Brian" == Brian Edmonds <edmonds@cs.ubc.ca> writes:

>>>>> "Steven" == Steven L. Baur <steve@miranova.com> writes:
Steven> I'd like to see retain an author copy on such lists only so long
Steven> as the message hasn't come back from the list server, then the
Steven> author copy can be replaced with the copy from the list.

Brian> Yeah, right.  Good luck determining this in any automated way.

[For the simple-minded case]
It costs some CPU cycles and disk storage, but the framework is
already implemented.  Set up a special spool area for copies of
outgoing mail-list mail, and make sure that your outgoing e-mail to
mailing lists gets archived there.  Then add something to your
incoming mail processing hooks to check message ids against those
saved in the spool area, deleting the message out of spool when a hit
(your message comes back) occurs.

If you maintained a table of Message IDs and disk file names, it might
even be reasonably efficient.

I agree there's no good general automatic algorithm to determine
whether the article returned is truly a faithful reproduction of the
original.  However, if you use a PGP signature and the function I
posted earlier to move some of the headers into the signed area, you
can test the returned signature, and compare the non-PGP signed
remainder against whatever extra text the list owner places on
outgoing messages.

-- 
steve@miranova.com baur

Re: Archived sent messages

[Lars Magne Ingebrigtsen]

Brian Edmonds <edmonds@cs.ubc.ca> writes:

> I'm not sure if the best approach for newbies is to have Lars' outgoing
> mail archive be the default, or to let the default be no outgoing
> archiving, as one would normally expect.  I lean towards the latter.

I honestly don't know what would be the most reasonable approach here.
If we archive automatically, then we make it more difficult to "lose"
mail -- people can always go back and find out what they said, which
might be nice.  But is it a good idea to promote hogging disk space
with lots of boring old mail and news as a default?

-- 
Home is where the cat is.

Re: Archived sent messages

[Per Abrahamsen]

I think the for other mailers is to create some sort of outgoing
message archive.

Re: Archived sent messages

[Lars Magne Ingebrigtsen]

Per Abrahamsen <abraham@dina.kvl.dk> writes:

> I think the default should be to let `nnmail-split-fancy' decide which
> group the outgoing messages go to.  Having in and outgoing messages in
> the same groups is really nice in a threaded mail reader, and I
> believe it would be a good default for new users.

Yes...  Or one could just Gcc to the current group, which might be
even more intuitive.  

But the problem is that most people do not want to expire the pearls
of wisdom they themselves have written... *ever*.  They're to be saved
until one gets famous and future biographers want something to dig
into.  (Or something.  :-)  

So I want my mail to be saved into one big file that I can gzip and
etch onto CDs somewhere periodically.  

-- 
Home is where the cat is.