Fox comment for declare-function.

[gnus] / todo

;; Also know as the "wish list".  Some are done. For the others, no
;; promise when to be implemented.

* Go through the todo list and remove items already done.

* Use a new custom type (`define-widget') for posting-style in `gnus-cus.el'
  (G c) and for `gnus-posting-styles'.  Maybe some allowed types are still
  missing.

* Add proper doc strings to functions and variables explained in the manual
  (info "(gnus)Gnus Utility Functions")

* Add Message-IDs or URLs refering to relevant discussions on lists and
  newsgroups.

* Use nicer tool bar icons from GNOME

  Done for Emacs (The GNOME icons won't fit into standard XEmacs icons,
  IMHO. -- rsteib) in group, summary and message mode.

  Some modes might also deserve improved tool bars:
  
  - gnus-draft-mode
  
  - mml-preview buffer:
  
    . zap most buttons; except print, customize (?) and help
  
    . "exit" should just kill the buffer
  
  - gnus-server-mode: Add some commands from the Connections and Server
    menu.
  
  - gnus-browse-mode (could borrow some icons from gnus-group-mode)
  
  (See http://article.gmane.org/gmane.emacs.gnus.general/62147).

* Maybe Gnus should support the LIST SUBSCRIPTIONS, see RFC 2980.

* Merge `message-extra-wide-headers' and ` message-header-synonyms'?

* Maybe texi/emacs-mime.texi could be divided into user-visible stuff and
  reference manual for the MIME library.

  Related: Bill Wohler's article on mh-e-user.
  http://thread.gmane.org/29067.1138078896@olgas.newt.com

* Fix `change servers' command, see David Kastrup's message.
  http://thread.gmane.org/x54qewqxz4.fsf@lola.goethe.zz

* texi/gnus-coding.texi should be fixed.

* gnus-topic-kill-region
  From Colin Marquardt <colin.marquardt@usa.alcatel.com>
  
  I noticed that when re-arranging topics, C-k yanks a topic just fine
  (runs gnus-topic-kill-group).

  However, my habit is to do marking and the yanking the region, so I
  would run C-w on the marked topic. But C-w runs
  gnus-group-kill-region and doesn't yank the topic (for groups it
  works fine).

  So could we have a gnus-topic-kill-region, or a
  gnus-group-kill-region which handles topics as well?

* Speed up sorting in summary buffer if there is a limit.
  
  Suggested by Daniel Ortmann <ortmann@isl.net>.

* Investigate the memory usage of Gnus. 

  But it does seem strange that Gnus would use some 15meg for this.  I
  think that is worth investigating.  I suspect that bugs or bad
  design are causing waste; they could be in Gnus, or in Emacs. -- RMS

* Google group digest
  
  The result of Google group search return a thread. Is it a digest
  format?

* NOV caching.
  
  Implement NOV caching with Gnus Agent.

* Allow specification of server in Newsgroups header
  
  [Kai wrote]

  WIBNI I could put `Newsgroups: nntp+quimby:bla' into a message and
  Gnus would know to post this message on my server `nntp:quimby' into
  the group bla?  I think this would be way cool.

  But Gnus would have to rewrite the Newsgroups header before actually
  sending the posting.

  Thanks for Micha Wiedenmann for this suggestion.

* Parsing of the common list confirmation requests so that Gnus can
   prepare the response with a single command.  Including LISTSERV
   periodic ping messages and the like.

* Parsing of the subscription notice to stash away details like what
   address you're subscribed to the list under (and automatically send
   mail to the list using that address, when you send mail inside the list
   group), what address to mail to unsubscribe, and the list info message
   if available.  Hitting the "get FAQ" command inside a mailing list
   group should display that stashed copy of the info message.

* Some help in coming up with good split rules for mailing lists, as
   automated as possible.  Splitting on To and Cc is almost always not
   what I want, since it can misfile messages and since if I'm cc'd on
   list mail I want to get both copies, one in my personal mailbox and one
   in the list mailbox.  I know other people handle it other ways, but I
   prefer it that way.  Accordingly, some way to semi-automatically
   generate split rules based on Sender, Mailing-List, Return-Path,
   X-Loop, and all of the other random headers that often work would be
   very cool.

* Support for zipped folders for all backends this makes sense for.
  Most likely using jka-compr. (It has been suggested that this do
  work but I think it should be verified for all backends.)

* Agent (Can someone write some subtopics here? I don't use it myself
  so I don't know what is lacking.)

* Support for encrypted folders. Even if the mail arrives unencrypted
  Gnus should be able to encrypt the *folder* for added safety. This
  should go for both Gnus' own folders and the folders Gnus reads from
  (e.g. /var/spool/mail/${USER}). All backends this makes sense for.

  [John Wiegley's article <200011030445.VAA08277@localhost.dynodns.net>,
   posted on gnu.emacs.gnus does this.
   Also, gnus-article-encrypt `K E' encrypts the article body.]

* Splitting .newsrc.eld so the history is in one file and the
  configuration is in another. To help those that reads at two
  locations (e.g. work and home) and want to have the same
  configuration.

* gnus-uu-decode should complain if one or more parts of a series post
  (ie, "part N of X") is missing, and optionally tick what parts are
  there for decoding in a later session.

* Additional article marking, and an ability to affect marks placed
  during e.g. mail acquisition.  I want to be able to notice the
  subject "fast money" or "web traffic", automatically mark it with a
  `$', and score it into oblivion.  (But I fear that wanting to change
  marks with mail-source-* and nnmail-* functions will represent a
  philosophical conflict with the rest of Gnus' management of article
  marks.  mail-source-* and nnmail-* currently hack around with files
  under ~/Mail and leave traces in ~/Mail/active, but don't affect
  things stored in .newsrc.eld.)

* A much better interface to nnmail-split-methods.  I don't know how
  I'd like this done, but I know that the current method of manually
  hacking regexps is pretty untenable for new users.  My boss, who is
  tenured faculty at CMU and CEO & CTO at JPRC, and whose research
  work has involved Lisp for the last 25 years, is trying to implant
  himself in a Gnus mail environment, and this is a big sticking point
  even for him.

* PGP-supported encryption of entire nnml & nnmh groups.  There are
  people with whom I exchange mail routinely who don't send w/PGP, but
  I'd really rather that the content not be left lying around
  unencrypted.  Hook into article acquisition the way jka-compr
  supposedly does, to auto-decrypt every message read.

  [See Support for encrypted folders.]

* Baby's First Mail In Gnus.  Some set of functions that the
  new-to-mail-in-Gnus user can invoke which will query the user
  appropriately for the basic information required to establish mail
  handling, leaving the appropriate traces in .gnus.  Perhaps a
  customize buffer would be appropriate.
  - Where does your mail come from?
  - If some server, what is your POP/IMAP protocol identity?
  - What is your identity when sending mail, as opposed to posting to
    Usenet?
  - Here are some basic concepts of mail groups (list a few:
    personal mail, company-wide mail, mailing lists, garbage dumps,
    receptacles for outbound copies of what one sends; which ones do
    you want to instantiate, and what mail should land in each?
    [/viz./ problem of nnmail-split-methods interface.]

  [Probably `assistant.el' will provide this.  But it's development is
  stalled.]

* Full integration of nnir into Gnus.  Generic hooks for adding new
  external nnir sources.  I use a couple experimental, in-house tools
  (JPRC is a research lab, occupied with document analysis and machine
  learning) and adding new search engines to nnir by hacking the main
  nnir.el module is rather clunky.

* Manual ordering of articles in an nnml folder.

  That is, keystrokes to move articles (or whole threads) up or down
  in the *Summary* buffer relative to the other articles.  The order
  would be persistent (e.g., across gnus sessions).

  With this ability, an nnml folder would make for a good to-do list.

* Since many uses Gnus to store to do lists I think it is time for an
  nntodo. (I know Kai already written one, maybe use that for a start?)

* nnsql backend, which would allow messages or folders to be imported
  in a local (My|Postgre|?)SQL RDBMS.

* "posting profiles" ideally accessible from a popup menu; allowing
  choice between predefined profiles of
  from,name,organization,etc. Example: I'm at home, but need to reply
  to a work mail; i can hit 'R', then use this command to switch to my
  'work' profile for purposes of this one reply. (This might already
  be possible with current Gnus, but I don't think so.)

* Better handling of the mail retrieving / splitting feature:
  - the variables <backend>-get-new-mail should not exist anymore. Mail
    retrieving should be a separate matter.
  - we should be able to split mails to groups AND backends at the same time.
  - meanwhile, we should still be able to associate certain mail sources with
    certain backends.

* A better interface to the agent download scoring rules, like the one
  for the other scoring rules.

* Editing of messages in the agents cache.

* More article marks (like '!' or '?').
  Maybe user defined marks that can be displayed as any choosen charakter,
  so one could do things like limiting on, to do whatever one likes with
  these articles.

* A possibility to add notes to messages. If those could include links
  to other (stored) messages this would be very practical.

* Allow article editing in groups which do not support it, but
  emulating it via deleting the old article and entering the new one
  into the group.  This would be very useful to support `T ^' (say) in
  nnimap groups.

* Allow user to specify which kinds of groups should be displayed.
  For example, I want to display all the groups that are displayed
  now, plus those which have cached messages in them.  (Gnus does
  display those with ticked messages but not those with
  cached-but-unticked ones.)  This would become even more important
  when we allow labels.

* Create new data type `article identifier' and use that instead of
  article numbers.  A first implementation could offer something like
  (num . 4711) but this could be extended.  This would be useful for
  using servers with *really* large numbers -- there we could have a
  bignum type.  It might also be useful for the nnweb and nnultimate
  thingies where article identifiers are not really numbers.

* Allow use of digests to keep related articles.  Normally, you use
  groups to group together articles which are thematically related.
  But sometimes, you have so many themes that this becomes
  impractical.  WIBNI I could have digests in a group, and there was a
  way to add a new article to one of the digests in that group?

  Or maybe what I really want is a way to tell Gnus that a specific
  thread should always be hidden (as in `T h') by default, while most
  other threads are not hidden by default.  Hm.

* New backend nnbabylfolder.  There is also nnbabyl which is like
  nnmbox but uses babyl format, but there is no babyl format
  equivalent of nnfolder.

* Make movement commands in summary buffer independent of `move after
  mark' behavior when marking articles.  Currently, if you don't want
  `E' to move to the next unread article, you have to set
  gnus-summary-goto-unread to nil, and then there is no way to move to
  the next or previous unread article.

  This one has two sub-tasks.  Providing the commands is one thing,
  finding out useful key bindings for them is another.  I think we
  could provide the commands first while not changing the behavior of
  the key bindings; then different people can experiment with
  different key binding schemes until we find something which suits
  many people.

* `Move to next/previous/first article' is a misnomer, since ticked
  articles are also unread but not moved to by these commands.  Should
  the terminology be fixed or the documentation, or what?

* Allow sorting of threads by newest article rather than by root of
  thread.  Consider the following thread structure:

    root1       Jan 1
      leaf1     Jan 4
    root2       Jan 2
      leaf2     Jan 3

  These two threads are sorted this way because root1 is older than
  root2.  I want an option to sort them the other way round because
  leaf1 is newer than leaf2.

* Improve editing of MIME messages.  I would like to use html-mode to
  edit the body of a text/html message, and enriched-mode for
  text/enriched messages, and so on.  This should go for multipart
  messages as well.  This is probably a hard one since Emacs currently
  does not allow several major modes per buffer.  But maybe it would
  be nice to hack Emacs to provide this infrastructure so that Gnus
  can make use of it?  This would also make it possible to provide
  nifty commands for editing the headers, for example, rather than
  relying on commands which do the same thing everywhere.
  message-x.el is really just a half-assed attempt at doing it, and
  while it is useful, that's not the way it should be done.

  I think Francisco Potort\e,Al\e(B already did something like this?

* Provide commands for editing MML tags.  For example, there could be
  a command mml-add-tag-attribute which prompts me for an attribute
  name (with completion, from the set filename, type, ...), and then
  for a value.  (This is like `C-c +' in psgml.)  Or there could be a
  command which showed me all the attributes in an MML tag and allows
  me to use TAB to move between them, and then to edit each attribute
  value.  (This is like `C-c C-a' in psgml.)

* Have Gnus automagically set group parameters for mailing list
  groups.  For example, if I have a splitting rule that automatically
  sorts ding@gnus.org into mail.ding, then Gnus should clue in, set
  the to-list parameter to 'ding@gnus.org', and set total-expire.
  (This is probably Hard (TM).  And of course the user should be able
  to configure what parameters exactly get set.)

* Along the same lines, automagically detect broken reply-to's.  (But
  don't auto-detect users legitimately setting a reply-to header that
  points back to the list.)

* Make it easier to change parameters on a set of groups,
  e.g. set/clear gcc-self on process-marked groups.

* Make it easier/possible to migrate between primary select-methods,
  if that concept is going to be kept.  Right now I have only one
  group on my primary server, and I'd kind of like to change from nntp
  to nnml, but apparently this doesn't work well.

* Make it possible to refer to uniquely-named groups without
  select-method prefix (e.g. mail.misc instead of nnml:mail.misc).

* Allow a user-defined picons directory for personal groups.

* Annotations as discussed last autumn. Be able to make comments to
  articles for all backends. The comments maybe should go into a
  seperate "backend", like nndraft.

* Catchup on a topic and all its subtopics. I.e. do "c y" when on a
  topic line in *Group*.

* Better/more advanced subject washing in *Summary*, see my
  js-gnus-simplify-subject-function I posted earlier this winter.

;; From Newest Features node. Some are not done.

* I would like the zombie-page to contain an URL to the source of the
latest version of gnus or some explanation on where to find it.

*  A way to continue editing the latest Message composition.

*  facep is not declared.

* Include a section in the manual on why the number of articles isn't
the same in the group buffer and on the SPC prompt.

*  Interacting with rmail fcc isn't easy.

*       Hypermail:
      http://www.uwsg.indiana.edu/hypermail/linux/kernel/9610/index.html

   [w3 or nnwarchive?]

* `^-- ' is made into - in LaTeX.

*  gnus-kill is much slower than it was in GNUS 4.1.3.

*  when expunging articles on low score, the sparse nodes keep
 hanging on?

*  starting the first time seems to hang Gnus on some systems.  Does
 NEWGROUPS answer too fast?

*  nndir doesn't read gzipped files.

*  when moving mail from a procmail spool to the crash-box, the
 crash-box is only appropriate to one specific group.

*  nnmh-be-safe means that crossposted articles will be marked as
 unread.

*  Orphan score entries don't show on "V t" score trace

*  when clearing out data, the cache data should also be reset.

* rewrite gnus-summary-limit-children to be non-recursive to avoid
exceeding lisp nesting on huge groups.

*  expunged articles are counted when computing scores.

*  ticked articles aren't easy to read in pick mode - `n' and stuff
 just skips past them.  Read articles are the same.