v0.8:
	@ signs are legitimate inside urls
	Support iconv #define's used on MacOS
	Renamed some files so filenames don't clash with case insentivity
	Don't compile network-enabled bits of mimelib
	Getopt.h is not required and unavailable on some platforms
	Solaris has no grep -q, use /dev/null
	Include fcntl.h for Solaris O_RDONLY
	Don't use TZ for my_timegm; unsetenv is not in Solaris
	Fsync files to disk for file-system transaction guarantees
	Don't use flock if unavailable, try fcntl instead
	Have better fallback behaviour for screwy charsets
	Crop anything that even smells like a control char in xml
	Include sys/types.h for off_t on BSD
	Fixed some minor -Wall warnings
	Tidied up the commit semantics
	Provide an option to import without power-failure safety
	Implemented a smart cache manager: lurker-prune
	Improved the man pages with helpful information
	Include a subject for emails without one
	Fixed title display bug
	Include UTF-8 as charset in response Content-Type
	Missed cassert in some files; fixes compile under g++ 3.3

v0.7:
	Made VERSION controlled by one place: configure.ac
	Fixed a bug where mailing list names were kept from Reply-To
	Unfold folded email headers for http; broke some attachments
	Don't include control chars in output xml; drop them
	Added mejo's manpage for lurker-index
	Open mode of 0666 instead of 0644; rely on the umask
	Support override for lurker.cgi location
	Install cgis to cgi-bin rather than docroot
	Replace list_host with archive
	Don't use symlinks; not portable and jam doesn't do it
	Support building against local mimelib for deployment without kde
	Compatability with older g++ 2.95 versions
	Support backwards reading in esort
	Fixed a bug with the empty string inserted into the database
	The database supports prefix partitioning
	Make search work like mindex
	Removed some debugging that was left in 0.6
	A search for no keywords now returns everything
	Added a '-d' option to save space on reimport
	Added 'lurker-regenerate' to port databases
	Truncate reply-to list to 50 messages
	Have a valid 'jump-to' date even with no messages
	Removed follow-ups; the tree already has them
	Give nice error messages for non-existant messages
	Don't reorder or prune chars from search terms
	Don't import the empty string
	Searches for keys with %, /, and + work

v0.6:
	Jam replaces automake + overkill autoconf madness; hooray!
	Mimelib replaces libc-client; hooray for a decent API!
	libesort replaces libkap -- should index faster
	Indexing is done atomically with lurker-index (command-line tool)
	lurker.cgi takes atomic database snapshots directly (no ipc)
	Basically, a reimplementation in C++
	

v0.1g:
	Ignore reply-to address if equal to mailing list address
	Fixes for things that gcc 3.2 correctly warns about
	Updated build files to newer automake/autoconf
	Fixed a redirection bug on older browsers with jump/*
	Fixed file descriptor hack that broke Solaris
	Fixed a segv on corrupt embedded rfc822 mime components

v0.1f:
	Fixed overlong browser caching of last mindex for a list
	Ignore SIGPIPE for cases where libst doesn't catch it
	Turn off the warning about broken email headers
	Corrected column formating
	Center the entire search selection

v0.1e:
	Don't warn if a maildrop adds whitespace before "From ..."
	Allow port numbers in URLs ("http://localhost:8080/foo")
	Report total number of messages on the front page
	Include an Expires http-equiv to match lurker expiry
	Output a Last-Modified and Date header from the lurker.cgi to match
		what the webserver outputs
	Removed an evil kludge that relies on C calling convention
	Removed exslt dependency so IE6 xsl works again
	Made output html w3c compliant
	Got mozilla xslt to work!
	Mangle mailto:s regardless of xf:escape-uri availability
	Support utf-8 mailto if xf:escape-uri available
	XSL support detection via javascript added
	Forward/backward links within a message added
	Display which replies are thread drift
	Prettier arrows from Kevin Teague
	Should compile on BSD again
	Fixed alignment issue on ia64
	Better decisions about how much to mmap
	Added a FAQ
	Added tail caching to libkap

v0.1d:
	Fixed an off by one error in btree.c (could assert-fail)
	Kap allowed wbuffers w/o append, which could consume a lot of RAM
		during the backup script
	Fixed a free(0) in mbox.c
	Quatered the cache given to a write buffer
	Quadrupled the cache dump frequency
	Added command-line option for disabling import
	Added command-line option for disabling service
	Made the splash page display only list names in four column mode
	Made the AVL trees reentrant
	Reintegrated the read buffer into libkap -> faster searches
	mmap is used for .append files to allow async writes
	Fallback when out of address space to read()/write() added
	We preallocate storage for append to prevent wasted writes
	Threads can span 3 months instead of 1
	No longer require numeric id tags for lists
	Mboxs can be given arbitrary strings for tags
	Automatically detect unconfigured mboxs
	Load config files more cleanly -- will facilitate config reloading
	Mailing lists are refered to by name in the UI instead of ID
	Include version information in pages
	Added command-line option to output import speed statistics
	Made error message about direct CGI invocation more informative
	URLs like www.debian.org with http:// are now non-relative
	Fixed the search algorithm which was broken since libkap read cache
		was implemented; now we get all hits again.
	Fixed a bug where adjacent list items in the expiry heap led to
		a SIGBUS error on swap

v0.1c:
	Added libkap which completely replaces all of libdb3 and flatfile
		-- this means the database needs to be reimported
		-- we can now import >6Gb email with reasonable speed
	Included a backup to text-file command
		-- allows migration for future database versions
		-- deals with cases of power-failure killing db
	Improved the regexps for email markup
	Added date-range keyword searching
	configure no longer tries to rebuild Makefile.in / aclocal.m4 / etc
	Japanese added
	Thorough unit tests for the database
	Much better conformance to strict ansi C
	Beautified XSLT ala Max
	Many smaller things that got lost in the frenzied switch to libkap

v0.1b:
	Improved installation documentation
	Configure allows dependencies to be in non root locations
	Fixed a configure problem on MacOS X
	Switched everything to ISO C99
	Added the write caching algorithm (about a 15x import speedup)
	Added a new database format for flatfile (about 2x speedup)
	Breaders can now reclaim boundary info memory
	Attempt to save the database on a segfault
	Profiling code added
	Minor bugfixes
	
v0.1a:
	Initial alpha release to test core lurker functionality.
