todo

mutt-1.5.12 源代码。linux 下邮件接受的工具。

Problems are listed in approximate order of priority.

- character set support: We should have a global cache of
  character to file name mappings.

- When displaying MIME headers, rfc 2047 decoding is applied (which
  should not happen), and rfc 2231 decoding is not applied (which
  should happen).

- Help formatting could be revamped a bit.

- re-add support for .mh_sequences files

- In the "attachment" menu, assume this:

	1 [text/plain, 7bit, 1.1K]           <no description>
	2 [message/rfc822, 7bit, 6.1K]       A test message
	3 [text/plain, 7bit, 0.1K]           |-><no description>
	4 [message/rfc822, base64, 2.5K]     |-><no description>
	5 [message/rfc822, base64, 2.7K]     `-><no description>

  (please note the "message/rfc822" attachments encoded as
  Base64; that's illegal, but Sun's Mailtool sends that
  kind of messages); then go to, say, attachment "4",
  delete it, and go to the main menu; you won't be able to
  quit the mailbox (ok, 'x' works, but 'q' doesn't).

  The problem here lies in the fact that mutt uses mailbox
  handling functions to access message/rfc822 type
  attachments.  We'd need to perform an additional
  decoding step before using these functions to fix this
  bug.

  Please note that mutt's just assuming RFC-compliant mail
  here.  Fixing this stuff may become a PITA.




- BODY struct should probably have a pointer to its
  corresponding HEADER struct.  this is needed for
  mh/maildir mailboxes so the correct pathname can be
  found.  Or perhaps all we need is a .hdr member of the
  STATE struct so that all of the MIME handlers can look
  up the corresponding HEADERs if need be?

- handle message/external-body in some fashion

- handle message/partial reconstruction

- make patterns generic (I have patches for this -tlr), and
  introduce generic menu limiting, menu pattern searching, and the
  like.  

  Note: This still requires some thought, since we'd have to store
  per-entry data in the menu structure.  As an alternative, we could
  extend the tag method to do something to more general flags. The
  latter approach would make the implementation of propper
  tag-prefix behaviour more simple: Functions should only be applied
  when a message is tagged and visible.  Additionally, we must not
  access a menu's max field directly any more: Adding an entry to a
  menu will require re-allocating and possibly updating the v2r
  array.  How do we handle "in-the-middle additions" properly?  Do
  they happen at all?

$Id: TODO,v 3.1 2005/02/12 19:29:52 roessler Exp $