source: trunk/doc/devel/TODO @ 141

$Id: TODO 140 2009-02-09 19:53:39Z tim $

If you are interested in contributing to this project, here's a few
things you could look into:

 - Currently there is no way on the command line to search for exotic 
   paths/types.  For instance, if reglookup encounters an unknown VK
   type, it just prints it out in Hex.  However, if you wanted to search
   specifically for that type, there is no way to do it.  Similarly, it
   isn't possible to specify certain binary or weird characters in
   paths.  Reglookup should take the user path and unquote each path
   component using the \xQQ syntax prior to searching.

 - Testing, testing, and more testing.  reglookup needs to be tested on 
   NT/XP/2k3/Vista.  A regression test suite would be nice too.  Some 
   thoughts on this include a script which randomly fuzzes an existing
   registry file, and tries to detect crashes of reglookup when parsing
   it.  Another test script might randomly truncate an existing registry
   file, which will help improve reglookup's parsing on fragmentary
   files.

 - Build system.  I do not wish to use automake/autoconf in this
   project.  I have also now decided that make is painful to use for
   everything.  I'd like to switch to a suite of shell scripts driven by
   minimal make rules.  If you got any ideas on this, shoot them my way.

 - Unicode support still needs improvement.  While parsing strings seems
   to be decent, UTF-8 output would be nice.

 - The interface between reglookup.c and regfi.c is much better than it
   used to be, but the parsing of data objects needs to be moved into the
   library.  The quoting syntax should stay in reglookup/reglookup-recover
   but the basic parsing of data types into proper structures should 
   happen in the library so that they are accessible to other users of the
   library.

 - NK/VK/SK record caching.  Right now, HBIN metadata and SK records are 
   cached, but it's pretty haphazard, and NK/VK records are repeatedly
   re-parsed.  A generic caching library has been introduced but needs to 
   be applied to NK records at a minimum.  Eventually, VK records and 
   data should also be cached separately and only be parsed when needed, 
   rather than when a key is loaded up front.  Caching also needs 
   configurable object limits, preferrably configurable at build-time.

 - It might be nice to have a way to filter results by security 
   descriptor information.  Maybe by MTIME as well.

 - Documentation.  The security descriptor output format needs to be 
   documented.  Also, function contracts should be added to the 
   lower-level functions of regfi.c.

 - The smb_deps.h and smb_deps.c content is almost eliminated.  Just need
   to integrate parts that are being kept into regfi or other modules.

 - Need to figure out a reasonably correct way to convert UTF-16LE charaters
   to ASCII under Windows/MingW or other platforms that don't have proper 
   libiconv support yet.  Then a build-time option or autodetection can 
   dictate which version of conversion function is used.

 - It appears the registry may actually support UTF-16LE names on keys, 
   if the key type field is set appropriately.  Once data parsing is 
   integrated into regfi, then the UTF-16LE handling routines (which 
   would then be built-in) should be used to properly handle this case.

 - Grep through the source for 'XXX', and you'll find more.