get crypt working for windows, maybe include with distribution - use
CryptEncrypt maybe?

tidy up Windows installation: 
     - use preprocessor definition to specify creation of single user 
       library
     - create InstallShield packages for each type of installation

update distribution

create collection distributions

tidy up macro files: 
     - fix text versions

Write a perl script to go through a translated macro file and generate
appropriate icons

create default collection images at build time with gimp script

convert crappy old scheme gimp scripts to perl

fix query caching - do caching in receptionist, finish off cross collection
searching, remove any remaining phrase-searching stuff in queryaction -
maybe wait for mgpp before worrying about this

create default images/macros for remaining Dublin core metadata elements -
maybe have a separate macro file for all these things to make it easier to
add to

get remaining plugins tidied up and using passed in options - file
extension option for all plugins

get all plugins to use multiread object and throw away the specialized GB
plugins

write a gzip/bzip/zip/tar plugin and take any remaining gzip stuff out of
rest of plugins

tidy up classifiers, make collection-specific ones more object oriented,
add section and compact options instead of using separate classifiers - get
classifiers using same option passing code as plugins now use

create a general classifier for doing stuff like NPepaList.pm does. should
handle things like HDLs organization classification, HDLs magazines section
etc.

sort out arrows/paging of browserclasses (for queryaction too!)
-DocumentBottomArrows?? - want arrows at bottom as well as top of pages for
gberg type collections

tidy up automatic help text - need to work out type of collection at build
time - get rid of the hacky HelpNoDocs, HelpBibDocs and HelpBookDocs config
file entries

fix up collection specific metadata - content negotiation??

append to init and error logs instead of overwriting

implement a more robust way of dealing with the way browsers encode form
arguments when you don't want them to - check that
_decodedcompressedoptions_ has been set up correctly everywhere

extend configuration file syntax - tidy up all the format stuff

replace DocumentImages, DocumentTitles and DocumentHeading with
DocumentIcon (values of true, false and formatstring - true may be part of
formatstring)

fix up formattools - bug preventing nested If/Or - check that no 'here'
pointers can run past 'end'

get numwords (and maybe other stats?) working in collection info - maybe
set as macros

tidy things up so that get_collectinfo is only called once per collection
(at init time for fastcgi) - it's currently being called all over the place

get BrowseFilter EndResults option to accept -1

add Language metadata to all documents by default??

fix the way the interface language preference is implemented - should check
somewhere that current language matches one of those specified in config
file

configuration:

	- finish off makefiles for windows, versions for servers, cd-roms
	  and cd-rom single users

	- check if litle/big endian

	- check if -static flag should be used??

	- add fastcgi to distribution

	- set up setup.bash etc during configuration (so GSDLHOME is full
	  path and script can be run from anywhere)

	- fix up the sed command in top level makefile - doesn't currently
	  work on the suns - this will be removed - no longer use
	  gsdlhome.h

	- check for crypt

	- autoconfigure to ask where cgi-bin directory is

bug when classification is empty. CL number still gets incremented so 
formatstrings don't match as expected

usage page

if document has only one level - e.g. TEXTPlug or HTMLPlug; top level type
should be Invisible so that tables of contents aren't displayed

add browserclasses to statusaction

add ability to use a separate formatstring for each index when displaying 
query results

rebuild remaining collections to use browsetools/fastcgi:

	- dlmail

	- ohist
	
	- whist

WebPlug - optimize to run reasonably quickly? - Currently works (I think)
but processes each frame within each combination of a frameset - this means
some html documents are processed hundreds of times and it takes forever to
build a small collection

look into possibility of clearing highlighting whenever you go to browse
etc.  i.e. highlighting only works when coming directly from a search

replace mk_all_dir function with perl module - see if speed improvements
are worthwhile

update collections - set up auto-updating where needed

        - csbib

	- tidbits - some of the import directory was deleted (oops..)

	- tcc - mail updating

add option to buildcol.pl to force documents to be paged/hierarchical. need
to use it for niupepa collection

sort documents between importing and building

fix plugins option to mkcol.pl

get polling for new collections to work when using fastcgi - should also
produce error message when unable to read collect dir (rather than just
exiting)

fix code to use 6 parameter version of gdbm_open() when running under
windows

fix util.pm to not use link() when running under windows

what to do with Makefiles for collection specific receptionists (like
cstr), use configure?