Hey Emacs, this is -*- outline -*- mode!

* ABI's to break:
** All result returns will be done as structs, not as XML. !!!
** Make sure that all results can be gotten in asynchronous mode (ie, avoid
   returning information in the blocking version as function arguments).
** Drop the support for finding out if an operation is pending.  After all, one
   or two more ways for a user to shoot themselves in the foot don't matter.
** Compatibility interfaces that can be removed in future versions:
*** gpgme_data_new_from_filepart
*** gpgme_data_new_from_file
*** gpgme_data_new_with_read_cb
*** gpgme_data_rewind
*** GPGME_Busy, GPGME_No_Request
*** GPGME_No_Passphrase
*** GPGME_Invalid_Recipient, GPGME_No_Recipients

* Thread support:
** Build thread modules for static linking (which just suck in the
   desired symbols the hard way). !!

* New features:
** notification system
   We need a simple notification system, probably a simple callback
   with a string and some optional arguments.  This is for example
   required to notify an application of a changed smartcard,  The
   application can then do whatever is required.  There are other
   usages too.  This notfication system should be independent of any
   contextes of course.
** --learn-code support
   This might be integrated with import. we still need to work out how
   to learn a card when gpg and gpgsm have support for smartcards.
** set_locale for thread safe and env independent locale selection.

* Documentation
** Add note about GPGME clearing out pointer return values.
** validity/trust

* Engines
** Do not create/destroy engines, but create engine and then reset it.
   Internally the reset operation still spawns a new engine process,
   but this can be replaced with a reset later.  Also, be very sure to
   release everything properly at a reset and at an error.
   Think hard about where to guarantee what (ie, what happens if start fails,
   are the fds unregistered immediately - i think so?)
** Optimize the case where a data object has an underlying fd we can pass
   directly to the engine.
** Move code common to all engines up from gpg to engine.
** engine operations can return General Error on unknown protocol
   (it's an internal error, as select_protocol checks already).
** When server mode is implemented properly, more care has to be taken to
    release all resources on error (for example to free assuan_cmd).

* Operations
** Passphrase callback should not copy password. !!!
** Export status handler need much more work. !!!
** Import should return a useful error when one happened.
*** Import does not take notice of NODATA status report.
*** When GPGSM does issue IMPORT_OK status reports, make sure to check for them
    in tests/gpgs   m/t-import.c.
** Genkey should return something more useful than General_Error.
** Factor out common code in _op_*_start functions.
** Add ATTR to return the number of subkeys or uids.
** Optimize the file descriptor list, so the number of open fds is
   always known easily.  This could replace the pending bit, too, with
   the exception of keylisting operations maybe.

* Error Values
** Map ASSUAN/GpgSM ERR error values in a better way than is done now. !!
** Verify (and document) if Read_Error, Write_Error, Pipe_Error set errno.

* Tests
** Write a fake gpg-agent so that we can supply known passphrases to
   gpgsm and setup the configuration files to use the agent.  Without
   this we are testing a currently running gpg-agent which is not a
   clever idea. !
** t-data
*** Test gpgme_data_release_and_get_mem.
*** Test gpgme_data_rewind for invalid types.
*** Test gpgme_data_read's readable feature.

* Debug
** Handle malloc and vasprintf errors.  But decide first if they should be
   ignored (and logged with 255?!), or really be assertions. !

* Build suite
** Make sure everything is cleaned correctly (esp. test area).

Bugs reported by Stephane Corthesy:
> In GpgmeRecipients, would it be possible to provide a function which  
> would return the validity assigned to a name contained in the  
> GpgmeRecipients instance?

> passphrase callback. If I use the same GpgmeContext as the one which  
> is currently asking for a passphrase, my app crashes: the r_hd in
> the  
> callback has become invalid; if I use a brand new one, the callback  
> is called recursively, when I ask to enumerate keys.