* Look at FEATURES packet to decide wether to use MDC.  We should
    rethink the whole preference implementation and drop that ugly hamehash
    stuff.  Is it possible to that in getkey?  We have to see whether we
    realy lookit up by user name or whethre there is a double lookup,
    first by userid and then later by keyid - which is bad.

  * options.skel wird mit umask von root installiert.

  * See how we can handle the symlinks in ./gcrypt

  * Add SIGSEGV handler to overcome zlib problems with truncated data.

  * Use --output for keylistings too.

  * Never allocate packet memory with a m-alloc, but use a specific function.

  * Should we change names like mpi_write in g10/ so that we don't
    use the prefix mpi here?

  * Implement the AXP syscall to enable bus traps for GLIB 2

  * parse a paramter file to do automatic key generation and to set
    parameters for the keygen menu.  Far easier than to do a dialog for
    each required parameter.

  * With option -i prompt before adding a key to the keyring and show some
    info what we are about to add.

  * Speed up calculation of key validation.

  * --disable-asm should still assemble _udiv_qrnnd when needed

  * Skip RO keyrings when importing a key.

  * Use the newest encryption key if only the main key has been given.

  * replace the keyserver stuff either by a call to a specialized
    utility and SOCKSify this utility.

  * Check the beginning of file to detect already compressed files (gzip,
    bzip2, xdelta and some picture formats)

  * Delay the read of the passphrase-fd afte a NEED_PASSPHRASE.  But this
    may break some scripts.

  * as soon as we have moved to KBX, we can check signatures at all places
  because there is no perfomance drawback as we can store the result of
  a verification in the KBX.  This enable us to better print information on
  revoked user IDs and signatures. Well, caching of non-self-signatures
  will still be complicated.


Nice to have
------------
  * Official test vectors for 3DES-EDE3
  * use DEL and ^H for erasing the previous character (util/ttyio.c).
    or better readline.
  * Print a warning if the directory mode is wrong.
  * Do a real fix for bug #7 or document that it is a PGP 5 error.
  * preferences of hash algorithms are not yet used.
  * Replace the SIGUSR1 stuff by semaphores to avoid loss of a signal.
    or use POSIX.4 realtime signals.  Overhaul the interface and the
    test program.  Use it with the test suite?
  * add test cases for invalid data (scrambled armor or other random data)
  * add checking of armor trailers
  * Burn the buffers used by fopen(), or use read(2). Does this
    really make sense?	And while we are at it:  implement a secure deletion
    stuff?
  * the pubkey encrypt functions should do some sanity checks.
  * dynload: implement the hint stuff.
  * "gpg filename.tar.gz.asc" should work like --verify (-sab).
  * for messages created with "-t", it might make sense to append the
    verification status of the message to the output (i.e. write something to
    the --output file and not only to stderr.
  * configure option where to find zlib
  * Display more validity information about the user IDs at certain places.
    We need a more general function to extract such kind of info from the
    trustdb.
  * Evaluate whether it make sense to replace the namehashs either by
    using the user ID directly or by using pointers into the trustdb.
  * --gen-prime may trigger a log_bug; should be a log_fatal.