Pulled fix from Adam Schobelock that removes timezone offset
	caching because it was creating incorrect behavor or Moscow.
	Bugfix from Sophie Hamilton: call tzset() when changing timezones.

	Bugfix: bail out of tests cleanly when tzset doesn't work.
	Bugfix: make sure $ampm is defined so no undefined string warnings
	are issued.

	Honor the GMT flag when parsing time specs.  Patch from
	kris at shannon.id.au.

	Bugfix: RT#31477 noon & midnight would match in the middle of strings.

	Added %v conversion for Time::CTime

	Added tests to disprove invalid bug reports.
	A bugfix for negative offsets that overflowed (from Vernon Lyon 
	<vlyon at uk.yahoo-inc.com>)

	Add a few more timezones (from Peter Ludemann <ludemann at yahoo-inc.com>
	and Alex Bowley <kilinrax at cpan.org>)

	Recognize fractional times counts like "1.5 minutes ago" 
	(from Simon Wistow <simon@thegestalt.org>)

	Recognize Dow "before last" (from Ole Craig <ocraig at stillsecure.com>)

	Change various __DATA__ to __END__ (from Nick Ing-Simmons 
	<nick at ing-simmons.net>)
	Various folks have complained that JulianDay isn't correct.  An
	errata section has been added to the documentation.

	Mark Ivey <mivey@r2tech.com> add '@' as a synonym for 'at'.
	Fixed typos.  

	Bryan Henderson <bryanh@giraffe-data.com> sent in a patch to
	allow NOW => 0 so that ParseDate can be used to parse relative

	ewan@baremetal.com sent in a patch adding a bit of ISO 8601

	David Alban <david@responsible.com> doesn't like allowing the
	hour 24:00.  I agree.  Gone if you have VALIDATE=>1.

	Still to be done: the support the large number of ISO 8601 
	date formats: http://hydracen.com/dx/iso8601.htm
	Mike Cramer <cramer@webkist.com> sent a patch to Timezone to
	handle changes in $ENV{TZ}. 

	Rani Pinchuk <rani@cpan.org> sent in a definition for the
	CEST timezone.

	On Phil.Moore@msdw.com's behalf the timezone 'MET DST' is now

	No real changes, just added a license.

	Paul Zimmer <pzimmer@bepress.com> found that '7/17/01' with
	PREFER_FUTURE failed badly.  Fixed.  This was also submitted
	by Dana Burd <DBurd@cpr.org>.

	Ben Daglish <bdaglish@yahoo.com> noted that array context
	was implying WHOLE and DATE_REQUIRED.  Fixed.

	Streppone Cosimo <cstrep@idgroup.it> requested support for
	postgresql's date format ("dd/mm/yyyy hh:mn:ss +tzo")

	Ameeth Linus Daniel <linus@teil.soft.net> found that IST wasn't 
	a defined timezone.  I've defined it as +0530, but I'm not 
	positive that's correct.

	Gautam Tripathi <dayalpura@hotmail.com> sent in a fix for
	timezone offset calculations making them consistent across
	all systems (no more problems with differing interpretations
	of negative modulus operations).

	Mike Nerone <mnerone@idworld.net> requested the syntax 
	"4 days ago".  He says that that syntax is suppored by
	GNU date.  This was also requested by Townsend, John E. 

	Anton Berezin <tobez@heechee.tobez.org> notice that ParseDate
	couldn't handle "Sunday, 06-Nov-94 08:49:37 GMT" and sent
	in a patch.

	W. Phillip Moore <Phil.Moore@msdw.com> reproted a Y2K bug: it could
	not parse 'Jan  1 2000 10:30:30AM'.  Fixed.  This is embarassing
	beasue <alangmead@boston.com> sent in a patch Dec 9th!
	Allen Smith <easmith@beatrice.rutgers.edu> noticed that VALIDATE
	wouldn't accept hours > 0...  
	Hugh C Kennedy <kennedyh@engin.umich.edu> contributed patches to
	parse months like `Jan.' in additon to `Jan' and `January'.
	Randy Kobes <randy@theoryx5.uwinnipeg.ca> sent in a fix for
	Timezone.pm that allows all tests to pass on MSWin32.
	Eric Prestemon <ecp@adknowledge.com> noticed that %y would print
	"0" in 2000, not "00" as it should.  Fixed.

	Parsedate() now checks for wantarray() and returns what remains of
	its input when called from array context.

	Added another parsing option: VALIDATE.  When VALIDATE is set,
	silly values for hours, months, etc will be rejected.  Like the
	32nd of December will no longer be new years day.
	Optionally return fractional seconds from ParseDate (including 
	the SYBASE formatted ones).  Provide for printing fractional
	seconds in CTime.  Changes from Douglas Wegscheid 
	More attempts to support systems whose time function isn't 
	centered around Jan 1, 1970 midnight GMT.
	Fixed a couple of compiler warnings.  Did extensive testing with
	dates beyond 2000 and before 1970.  Quite a few code changes.
	Old code would have worked in many > 2000 cases but broke 
	terribly for dates < 1970.
	New code should work until year 10k.  Well, there's one problem:
	unix time format is only 32 bits, signed. 
	Slight change to avoid an "Use of uninitialized value" error
	pointed out by Mike Coffin <mhc@Eng.Sun.COM>.
	It turns out that Sybase emits time values like "3:09:59:000PM".  The
	last three digits are milliseconds.  Such times are now recognized
	but the milliseconds are ignored.  Change requested by
	ryanmcleish <ryan@email.labmed.umn.edu>.

	Corrected the documenation on strftime conversions for %d and %e.  The
	code and the documentation now agree:
		%d - 01 to 31
		%e -  1 to 31
	Minor bugfix: didn't recognize 5/18/1998.
	Fix a 5.004 warning.
	Add support for "June 4, 1997"
	Added support for negative relative times.
	Bugfix: "now" by itself was ignored.
	Bugfix: could not combine "now" with time offsets (only date offsets)
	Added support for "TZN
	This is the first release to use the Makefile.PL 
	packaging.  There are many changes with repect to 
	previous releases.  All modules now running with -w and
	use strict.

	CTime.pm: should be faster
	ParseDate.pm: core logic changed, more dates parsed.
	datetime.t: more tests
	Timezone.pm: ripped out tzset code and replaced it
		with code from Graham Barr <bodg@tiuk.ti.com>
	Patch for UK-style dates: Sam Yates <syates@maths.adelaide.edu.au>