== Revision History for Mail::IMAPClient
Changes from 3.17_01 to ?       made by Phil Pearl (Lobbes)
Changes from 2.99_01 to 3.16    made by Mark Overmeer
Changes from 0.09    to 2.99_01 made by David Kernen
	- Potential compatibility issues from 3.17+ highlighted with '*'

version 3.43: Tue Feb 16 16:53:26 UTC 2021
	- admin: update copyright, move from rt.cpan.org (retired) to github
	- rt.cpan.org#132720: compress: Readmethod: avoid false-positive EOF
	  [Eric Wong]
	- rt.cpan.org#129920: Parsing headers fails under special condition
	  [Jens Berthold]
	- rt.cpan.org#132654: avoid cyclic reference on ->compress
	  [Eric Wong]
	- rt.cpan.org#132380: append_string can fail to match APPENDUID in response
	  [Radomír Polách]
	- rt.cpan.org#133496: allow for undocumented BODYSTRUCTURE extensions
	  [Rhesa Rozendaal]

version 3.42: Sun Feb 24 00:43:29 UTC 2019
	- rt.cpan.org#12859: has_capability() changes in 3.41 broke imap4rev1()
	  [Gilles Lamiral and Gábor Leszlauer]
	- updated http:// to https:// URLs for referenced resources
	- updated copyright for 2019

version 3.41: Thu Feb 21 01:47:42 UTC 2019
	- rt.cpan.org#128220: unseen(), messages() and related POD cleanup
	  [Dan Jacobson]
	- rt.cpan.org#128264: parse_message() minor code/POD cleanup
	  [Dan Jacobson]
	- rt.cpan.org#128215: verb missing in messages() POD
	  [Gilles Lamiral and Dan Jacobson]
	- rt.cpan.org#127271: simplify capability handling via has_capability enhancements
	  +* has_capability() success returns(true) with server response data, not always '1'
	  + enhanced t/capability.t test cases
	  + updated POD for capability() and has_capability()
	  [Gilles Lamiral and Mark Overmeer]
	- rt.cpan.org#122373: use of IO::Socket::IP led to connect(empty args) regression
	  [Gilles Lamiral]
	- rt.cpan.org#128127: fix minor POD typo for search()
	  [Gregor Herrmann]
	- folders()/subscribed() remove mailboxes with \Noselect attribute
	  [Ashley Willis]
	- fetch_hash() remove quotes around header names (seen w/outlook.com)
	  [Ashley Willis]
	- use first over grep for minor efficiency gains
	- other minor POD cleanup

version 3.40: Thu Dec  6 01:44:16 UTC 2018
	- rt.cpan.org#122373: support IPv6 by using IO::Socket::IP over IO::Socket::INET
	  [Gilles Lamiral and Mark Overmeer]
	- rt.cpan.org#127103: flags() undef value as an ARRAY reference on a bogus message
	  [Gilles Lamiral]
	- rt.cpan.org#124523: update examples/populate_mailbox.pl timegm usage
	  [Bernhard M. W.]
	- t/capability.t: added first set of tests
	- t/quota.t: minor fix when tests skipped

version 3.39: Fri Feb  3 00:43:00 UTC 2017
	- rt.cpan.org#115726: uninitialized value via fetch_hash
	  [Malte Stretz]
	- rt.cpan.org#119523: better error reporting on failed TLS connections
	  [Matthew Horsfall]
	- rt.cpan.org#114904: document noop()
	  [Glenn Golden]
	- rt.cpan.org#97718: (redux) never retry DONE
	  [Laurence Darby]
	- _imap_command() new doretry => 0|1 option to suppress/allow retry
	- updated copyright for 2017

version 3.38: Tue Feb  9 02:48:21 UTC 2016
	- rt.cpan.org#107592: redact credentials via debug if !Showcredentials
	  [Gilles Lamiral]
	- rt.cpan.org#110273: failure to quote password values
	  (regression introduced in 3.36 via fix for rt.cpan.org#100601)
	  [Gilles Lamiral]
	- rt.cpan.org#107593: allow getquota("")
	- *Quote() now returns qq("") for defined but empty values
	- rt.cpan.org#107011: fix folders_hash() docs and usage in is_parent()
	  [Gilles Lamiral]
	- rt.cpan.org#106500: split UID EXPUNGE with a large sequence set
	- added t/quota.t and t/lib/MyTest.pm for testing
	- updated copyright for 2016

version 3.37: Fri Aug 14 11:04:53 EDT 2015
	- regex fix in rt.cpan.org#96575 required Perl 5.10
	    updated regex to be backwards compatible with Perl 5.8
	  [Paul Howarth]

version 3.36: Fri Aug 14 01:40:50 UTC 2015
	- rt.cpan.org#96575: parsing fixes for fetch_hash and bodystructure
	  + better handling of data containing escaped characters
	  [Ken Neighbors]
	- rt.cpan.org#97718: hang waiting for tag if DONE sent outside of IDLE
	  [Laurence Darby]
	- rt.cpan.org#100683: failed literal commands hang waiting for '+'
	  [Matthias Pitzl]
	- rt.cpan.org#100601: login() fails if username requires use of literals
	  [Andy Lyttle]
	- *rt.cpan.org#103823: is_parent() should return undef on \NoInferiors
	  [Gilles Lamiral]
	- rt.cpan.org#105456: fetch_hash fails if sequence number contains '*'
	  + *no longer removing unrequested data (except UID) in returned hashes
	  [Gilles Lamiral]
	- rt.cpan.org#91912: selectable broke in 3.29 due to \b around \NoSelect
	  [Justin Vallon, Gilles Lamiral]
	- some documentation, test cleanup and new tests

version 3.35: Fri, Nov 22, 2013  2:18:41 PM
	- *use Quote() over Massage() to avoid stripping double quotes from arg
	  + removed constant NonFolderArg
	  + may deprecate Massage() in a future release
	- _list_or_lsub failed to handle folder names with double quotes
	  + cleanup Escaped_results()
	  + new methods: Escaped_history() _escaped_trans_data()
	  + removed method: _list_response_preprocess()
	  [Gilles Lamiral, Victor Lé]
	- rt.cpan.org#90315: getacl on a folder returned as literal failed
	  + _read_line could remove extra space when literal data was received
	  [Carl Karcher]
	- rt.cpan.org#89040: typo fixes
	  [David Steinbrunner]

version 3.34: Fri, Sep 27, 2013 12:50:17 AM
	- make Makefile.PL use non-interactive and document test.txt usage
	- new attribute: Socketargs => [ (IO::Socket::.. args) ]
	  + cleanup connect() to more flexible with IO::Socket::* args
	- untagged server data during send literal may cause client to hang
	  [Arthur Wolfe, Josh Hillman]
	  + _send_line() needs '+' only to know it is OK to send LITERAL data
	  + created _response_code_sub() to simplify _get_response()
	- remove internal "Folders" cache
	- Allow for RFC 6154 "IMAP LIST Extension for Special-Use Mailboxes"
	  [Mathias Reitinger]
	  + new method: folders_hash()
	  + deprecate: xlist_folders(), xlist()

version 3.33: Tue, May 14, 2013 10:12:43 AM
	- more cleanup on use of $@ and $!
	- cleanup get_bodystructure / get_envelope
	- allow Ssl arg as an arrayref to pass args to IO::Socket::SSL
	  [Ramana V Mokkapati]
	- no need to Massage() the folder name in uidnext()
	- rt.cpan.org#84028: get_envelope() fails when subject ends w/backslash
	  [Andy Lyttle]
	- rt.cpan.org#79476: move()/copy() with sequence causes numeric warning
	  [Oleg G]
	- *move()/copy() no longer sort message(s) provided by caller

version 3.32: Fri, Aug 10, 2012	 4:43:24 PM
	- document RFC2087 quota related calls
	  [Mathias Reitinger] documentation request
	- rt.cpan.org#78474: idle/idle_data documentation error
	  [Dima Kogan]
	- Quote()/Massage() now uses literals for non ascii data
	  [Mathias Reitinger] reported issues with utf8 data in password
	- use Quote()/Massage() consistently now in:
	    login() proxyauth() deleteacl() setacl() listrights() rename()
	- documented deleteacl() and other minor pod cleanup
	- ran Mail::IMAPClient::BodyStructure through perltidy
	- update year in README/pod to 2012
	- rt.cpan.org#74733: Fails with Parse::RecDescent >= 1.966_002
	  rt.cpan.org#74593: Recent changes break Module::ExtractUse and ...
	  [ANDK, TEAM, SREZIC, NBEBOUT at CPAN and nine from detonation]
	  - Makefile.PL avoid buggy Parse::RecDescent 1.966_002 until 1.967_009
	- rt.cpan.org#76989: Mail::IMAPClient::BodyStructure usage/docs
	  [Pierluigi Frullani]
	  - fix incorrect documentation on new()
	  - lots of doc verbiage updates

version 3.31: Mon, Mar 19, 2012 11:11:11 AM
	- rt.cpan.org#74799: Support for partial data responses in fetch_hash
	  [Philip Garrett]
	  + bonus: cleaner handling of BODY.PEEK responses
	- properly handle ALL|FULL|FAST fetch macros in fetch_hash

version 3.30: Fri Nov 11 09:37:00 EST 2011
	- rt.cpan.org#72347: Starttls array ref argument dereferenced twice
	  [Jonathan Buhacoff]
	- during connect(): Port now defaults 143 or 993 if $self->Ssl
	  [Kodi Arfer]
	- stop reconnect deep recursion if server disconnects on login
	  [Luca Ferrario]
	- reconnect() now returns 1 on success; on error undef or 0=recursive
	- handle EBADF from syswrite in _send_bytes
	- rt.cpan.org#67263: add RFC4978 IMAP COMPRESS Extension support
	  + new method: compress()
	  + new attributes: Compress Readmoremethod
	- general code cleanup:
	  + new() now always returns $self or undef (never $sock any more)
	  + Socket() now always return a socket or undef
	  + login() now always return $self or undef
	  + _read_more() will now use Readmoremethod if set
	- missing second arg '' for encode_base64 causing AUTHENTICATE
	  PLAIN to fail on lines longer than 76 characters
	  [Yoshiho Yoshida]

version 3.29: Tue Aug  9 00:33:52 EDT 2011
	- rt.cpan.org#69876: ENVELOPE as part of fetch_hash convenience method
	  [Chris Huttman]
	  + added Mail::IMAPClient::BodyStructure::Envelope->parse_string($str)
	    convenience method for handling ENVELOPE data from fetch_hash
	- rt.cpan.org#68310: folders() should not call exists()/STATUS
	  [Gilles Lamiral]
	    - affects folders() and subscribed() methods
	  + use selectable() instead of exists() in call
	    - consider removing extra call to folders()/subscribed()
	  + ensure separator is set properly in folders()
	  + selectable now properly checks for \Noselect flag
	  + update folders() POD to match implementation behavior
	- rt.cpan.org#68648: [patch]: CAPABILITY after authenticate
	  [Stef Simoens]
	  + delete cache after State set to Authenticate
	    - State() is no longer an auto-generated method
	- rt.cpan.org#68755: provided socket loses blocking in 3.19-3.28
	  [Martin Schmitt]

version 3.28: Fri Mar  4 00:17:38 EST 2011
	- rt.cpan.org#66004: internaldate() return undef if no internaldate in reply
	  [Jason Long]
	- rt.cpan.org#66367: fetch_hash uses Escaped_results() in 3.26/3.27
	  (redo) rt.cpan.org#63524: fetch_hash() parse errors
	  [Mathias Reitinger]
	  + fetch_hash: only Escape() data in parenthesized list
	  + update fetch_hash test and add a new test
	- do not touch CRLF in Escape()/Unescape()
	- added Escape() method
	- rt.cpan.org#66287: flags results truncated due to Maxcommandlength
	  [Erik Colson]
	- rt.cpan.org#65694: SASL PLAIN: bad order of login data
	  [Willi Mann]

version 3.27: Sun Feb 13 14:37:27 EST 2011
	- rt.cpan.org#65694: migrate fails
	  [Erik Colson]
	- rt.cpan.org#65470: uninitialized warning in message_to_file
	  [Gilles Lamiral, Mark Hedges]
	- rt.cpan.org#61835: (DOC) in LIST context undef may be returned
	  [Stefan Völkel]
	  + warn/highlight behavior in docs Errors section
	- updated documentation
	  + migrate() documentation fixed
	  + moved Custom Authentication Mechanisms toward end
	  + recommended use of scalar context due to historical API behavior

version 3.26: Mon Jan 31 22:15:04 EST 2011
	- *require Perl 5.8.1 as constant use is invalid on 5.6
	- rt.cpan.org#63524: fetch_hash() parse errors
	  [Brian Kroth]
	  + fixed handling of LITERAL values in response
	  + fixed handling of field names with a dash (e.g. X-SAVEDATE)
	  + fetch_hash now uses Escaped_results() method
	- *fixed Escaped_results() to properly join LITERAL data with
	  the data that comes before and after it
	- *rt.cpan.org#60945: append_file() does not interpret $date as expected
	  [Jason Long]
	  $date should now be 1 (to use the file mtime) or a valid RFC3501 date
	- *rt.cpan.org#61292: memory consumption with message_string()/append()
	   rt.cpan.org#61806: Major problem with one function in IMAPClient
	  [Gilles Lamiral, Casey Duquette]
	  + use @_ / $_[<num>] in critical places to avoid pass by
	    value memory overhead
	  + use in memory files in a few critical places as that code
	    path in Mail::IMAPClient is significantly more efficient
	    with internal memory usage
	  + *new (undocumented/do-not-use-without-good-reason)
	    attribute Maxappendstringlength used by append() and
	    append_string() holds the size (in bytes, default 1 MiB)
	    that triggers when message SCALAR(s) passed to these
	    methods will be treated as an in memory file.  This
	    attribute will likely be removed in a future version.
	  + *append() and append_string() now call append_file() and
	    use an im memory file when length($message) is greater
	    than Maxappendstringlength; other minor code cleanup
	  + *message_string() now calls message_to_file() and uses an
	    in memory file
	  + refactor message_to_file() to use internal _imap_uid_command()
	  + update _read_line() to be more efficient w/CPU in critical
	    section by pulling isa() checks out of main loop also
	    conserve memory by not storing an extra copy of LITERAL
	    data if the data was put into a filehandle from the caller
	  + Memory/working set (KB) comparison (Perl 5.10 cygwin Win7):
	    - test: message_string on 6.1M msg and then append 6.1M msg
	      version | start | after message_string | after append
	        2.2.9 |  7624 |                74404 |       131896
	         3.25 |  7716 |                74408 |       156532
	         3.26 |  7684 |                33372 |        42608
	- minor arg cleanup of noop() and tag_and_run()
	- rt.cpan.org#63444: relax get_envelope(), allow empty reply-to
	  [Nikolay Kravchenko]
	- rt.cpan.org#61068: append_string can invalidate a good $date
	- rt.cpan.org#60045: Logout error if delay between BYE and tagged OK
	  [Armin Wolfermann]
	  no longer set an error when this happens
	- rt.cpan.org#61062: migrate() errors
	  [Johan Ekenberg]
	  + rewrote migrate() to be functional and simple
	- Update README and cleanup several old or out of date files

version 3.25: Fri May 28 00:07:40 EDT 2010
	- fix body_string parsing bug and added tests in t/body_string.t
	  [Heiko Schlittermann] 
	- rt.cpan.org#57661: uninitialized value warning in IMAPClient::thread
	  [Max Bowsher]
	- rt.cpan.org#57337: Correctly handle multiparts in BodyStructure.pm
	  [Robert Norris]
	  fixes in Mail::IMAPClient::BodyStructure::bodystructure for
	  bugs still in release 3.24
	- rt.cpan.org#57659: install fails when using cPanel GUI
	  [Ken Parisi]
	  hack Makefile.PL to use alarm() and timeout prompt() gracefully
	- relax t/basic.t logout() error check (allow 'BYE' instead of 'OK')
	- left examples/idle.pl out of MANIFEST for 3.24

version 3.24: Fri May  7 17:02:35 EDT 2010
	- rt.cpan.org#48912: wrong part numbers in multipart messages
	  [Dmitry Bigunyak, Gabor Leszlauer]
	- fix Mail::IMAPClient::BodyStructure::bodystructure to
	  properly assign parts for messages using multipart and also
	  include .TEXT parts as well (still not including top level
	  HEADER and TEXT though - bug?)
	- allow _load_module() to set $@ and LastError if module load fails
	- rt.cpan.org#55527: [no] disconnect during DESTROY
	  [Stefan Seifert]
	- updated logout documentation to correctly state that DESTROY
	  is not used to force an automatic logout on DESTROY despite
	  documentation that indicated otherwise
	- update append* documentation to match current implementation
	- rt.cpan.org#55898: append_file can send too many bytes
	  [Jeremy Robst]
	- avoid append_file corner cases operating on lines instead of buffers
	- use binmode on filehandle in append_file
	- add tests to t/basic.t for append_file
	- rt.cpan.org#57048: _quote_search() using $_ in loop instead of $v
	  [Matthaus Kiem]
	- added examples/idle.pl program showing use of idle and idle_data
	- idle_data() should not read/block after server returns data
	  [Marc Thielemann]
	- idle_data() _get_response regexp updated to not match errors
	- idle_data() now uses a timeout of 0 by default as documented
	- _get_response() now checks for defined($code) to allow $code==0

version 3.23: Fri Jan 29 00:39:27 EST 2010
	- new beta idle_data() method to retrieve untagged messages during idle
	  similar to method suggested by Daniel Richard G
	- added/updated documentation for idle, idle_data, and done
	- rt.cpan.org#53998: fix NTLM auth: call ntlm with challenge string
	  [Dragoslav Mlakar]
	- report the return value from select/_read_more on errors
	- logout() again returns the success/failure of the LOGOUT command
	- set/return error when $response->() returns undef in authenticate()
	- new internal method _load_module() centralizing some 'require' calls
	- localize use $@ in several places to avoid stomping on global val
	- refactor code calling _read_more() to centralize error handling

version 3.22: Thu Jan 21 15:25:54 EST 2010
	- rt.cpan.org#52313: Getting read errors if Fast_io is set to 1
	  [Jukka Huhta]
	- updated Maxttemperrors docs related to EAGAIN handling
	- new starttls() method and Starttls attribute to support STARTTLS
	- update parse_headers to try harder to find UID in fetch response

version 3.21: Tue Sep 22 19:45:13 EDT 2009
	- rt.cpan.org#49691: rewrite of fetch_hash to resolve several issues
	  [Robert Norris]
	  includes new tests via t/fetch_hash.t
	- rt.cpan.org#48980: (enhancement) add support for XLIST extension
	  [Robert Norris]
	- rt.cpan.org#49024: NIL personal name returned by *_addresses methods
	  [Dmitry Bigunyak]
	- rt.cpan.org#49401: IMAPClient expunge fails (unless folder arg used)
	  [Gary Baluha]
	- update/clarify close and expunge documentation a little

version 3.20: Fri Aug 21 17:40:40 EDT 2009
	- added file/tests in t/simple.t
	- added methods Rfc3501_date/Rfc3501_datetime
	  used by deprecated methods Rfc2060_date/Rfc2060_datetime
	  rt.cpan.org#48510: Rfc3501_date/Rfc3501_datetime methods do
	    not exist [sedmonds]
	- login() hack to quote an empty password
	  rt.cpan.org#48107: Cannot LOGIN with empty password [skunk]

version 3.19: Fri Jun 19 14:59:15 EDT 2009
	- *search() backwards compat: caller must quote single arg properly
	  rt.cpan.org#47044: $imap->search does not return [ekuemmer]
	- cleanup regexp in _send_line()
	- reduce extra newlines injected by _debug()

version 3.19_02: Tue Jun  9 00:47:52 EDT 2009
	- _list_or_lsub() now calls _list_response_preprocess so
	  consumers of this method no longer need to deal with how
	  LITERAL data is represented in the returned data
	- update _list_or_lsub_response_parse handling of folder names
	  that came back as literal data
	- update comments related to _list_response_preprocess
version 3.19_01: Fri Jun  5 15:45:05 EDT 2009
	- make parse_headers more robust to errors/non-header data

version 3.18: Wed Jun  3 23:07:12 EDT 2009
	- enhance fetch_hash to enable caller to specify list of messages
	  suggestion by [Eugene Mamaev]
	- better handling of untagged BYE response

version 3.18_02: Wed May 27 10:02:24 EDT 2009
	- *new attribute Ssl, when true causes IO::Socket::SSL to be
	  used instead of IO::Socket::INET.  This change allows
	  Reconnectretry logic to work on SSL connections too.
	- have LastError cluck() if setting error to NO not connected
	- handle errors from imap4rev1() in multiple places
	- Reconnectretry/_imap_command enhancements/fixes
	  + only run command if IsConnected
	  + keep a temporary history of LastError(s)
	  + sets LastError to NO not connected if ! IsConnected
	  + retry =~ timeout|socket closed|* BYE| NO not connected
	- _imap_command_do reduce data logged when using APPEND
	- fetch() now handles messages() errors
	- thread(), has_capability(), capability() better error checking
	- authenticate() now uses _imap_command for retry mechanism
	- size() now sets LastError when no RFC822.SIZE is found 

version 3.18_01: Fri May 22 17:08:00 EDT 2009
	- *update several methods to use common _get_response() method
	- refactor most code handling imap responses
	- new internal method _get_response() to reduce code duplication
	- more regex cleanup $CR/$LF (not \r\n) per perlport/IMAP spec
	- major cleanup/fix of append_file for rt.cpan.org#42434

version 3.17: Thu May 21 01:40:08 EDT 2009
	- ran all test code and lib/Mail/IMAPClient.pm through Perl::Tidy
	- plan on using perltidy to standardize format going forward
	- added 13 tests to t/basic.t to cover more methods
	- fix some broken tests
	- update Makefile.PL to provide info about optional modules

version 3.17_05: Tue May 19 11:04:28 EDT 2009
	- *reset LastError for every call to _imap_command_do()
	- *run() - use _imap_command_do(), return arrayref in scalar context
	- *tag_and_run() - return arrayref in scalar context
	- *done() - use _imap_command_do(), return arrayref in scalar context
	- *search() now returns empty arrayref not undef if no matches found
	- _imap_command_do() made more flexible to avoid code duplication
	- _list_response_parse renamed _list_or_lsub_response_parse
	- updated POD with new/updated behavior
	- append_string() now uses _imap_command_do() for Reconnectretry
	- internally use defined return values instead of only LastError()
	- run() updated to use same/similar code to _imap_command_do()
	- make several return statements more consistent
	- delete() now unsets current Folder attribute on success

version 3.17_04: Fri May 15 17:18:52 EDT 2009
	- updated POD with new reconnect() method and Reconnectretry attr
	- *new _imap_command() after renaming old one to _imap_command_do
	  support retrying commands X times EPIPE/ECONNRESET errors
	- *new Reconnectretry attribute to control number of retry
	  attempts (default is 0 - no reconnect/retry)
	- *added reconnect() method to support Reconnectretry attr
	  reconnect and updated _imap_command() method
	- *_imap_command_do will return undef if command given has no TAG
	- fixed message_string() logic/errors for failed size() calls
	- local-ize $! anywhere we use Carp routines as older versions
	  of Carp could cause $! to be reset
	- several 'BUG?' comments -- raising red flag for future work
	- minor cleanup of sort() logic
	- reduce duplicate code, hopefully improved error handling:
	  new _list_or_lsub() for list() and lsub()
	  new _folders_or_subscribed() for folders() and subscribed()
	  + new _list_response_preprocess() keeping old code/logic in
	    for now, but may remove in the future (for buggy servers?)
	- some updates for migrate() but this method needs much work
	- body_string() now handles fetch() errors
	- tag_and_run now handles _imap_command() errors
	- changed non-timeout CORE::select() timeout from 0.001 to 0.025
	- minor cleanup of _read_line() error handling/debug output
	- get_bodystructure() handle more fetch() errors
	- expunge() handle select() errors
	- restore_message() handle store() errors
	- uidvalidity() handle status() errors
	- uidnext() handle status() errors
	- is_parent() use _list_response_preprocess() for parsing
	- move() send delete_message() errors to stderr
	- simplify size() method

version 3.17_03: Fri May  8 16:37:08 EDT 2009
	- *added uidexpunge() for UID EXPUNGE UIDPLUS support
	- *search() now DWIM: auto-escapes args, SCALAR refs not escaped
	  rt.cpan.org#44936 [cjhenck]
	- _quote_search() provides auto-escape capability for search()
	- many POD updates as well as some major reformatting (incomplete)
	- login now fails if passwd and user are not defined
	- _sysread(): $self was in args to 'Readmethod' twice
	- authenticate() return undef on scheme eq "" or LOGIN
	- "require" instead "use" Digest::HMAC_MD5 for CRAM-MD5 support

version 3.17_02: Fri May  1 16:44:21 EDT 2009
	- cleanup of use/imported data
	- use Socket $CRLF in many cases not \r\n per perlport/IMAP spec
	- *new Keepalive attribute used via new()/Socket() enables SO_KEEPALIVE
	- LastError now uses Carp::confess for stack trace if Debug is true
	- Maxcommandlength now defaults to 1000 per RFC2683 section
	- added noop() to support IMAP NOOP
	- _imap_command now sets LastError if a OK/$good response is not seen
	- fixed fetch_hash() to return FLAGS as "" not () when no FLAGS set

version 3.17_01: Fri Apr 24 18:36:45 EDT 2009
	- *new attribute Maxcommandlength used by fetch() to limit
	  length of commands sent to a server.  This should removes
	  need for utilities like imapsync to create their own split()
	  functions and instead allows Mail::IMAPClient to hopefully
	  "do the right thing"
	- remove extra 'use' calls for Carp and Data::Dumper
	- _read_more() improperly initialized vector causing select
	  errors, thus timeouts were not working properly (now they
	- *change default timeout 30s => 600s: 30s seems too short in
	- *explicit import of encode_base64 and decode_base64 from
	  MIME::Base64 note the code forces a disconnect from the
	  server on timeout as we can not easily recover from this
	  situation right now in the code
	- *numerous changes of error messages, removing superfluous
	  text and now relying on LastError instead of $! or $@ when
	- separator():
	  + now return undef if an error occured for NAMESPACE or LIST calls
	  + *no longer defaults to '/' if NAMESPACE call does not succeed
	- new internal _list_response_parse() method for parsing LIST
	- handle ECONNRESET errors on syswrite and mark connection as
	  + error "Connection lost" changed to "Write failed"
	- previously untrapped syswrite error now generate "Write
	  failed" errors
	- fix in _imap_command where LastError would be erroneously
	  set on LOGOUT
	- _record() no longer tries to infer errors based on data
	  being "recorded"
	- _send_line()
	  + cleanup in watching for: +|NO|BAD|BYE
	  + now sets LastError when an unexpected response is seen
	- _read_line()
	  + handle select errors instead of ignoring them
	  + forcefully _disconnect() on timeouts as this breaks app logic
	  + reduced duplication of code on error handling
	- added _disconnect() method to brute force drop connections
	  on timeout
	- added _list_response_parse() to reduce duplicate code for
	  LIST parsing
	- added _split_sequence() to support new Maxcommandlength argument
	- fetch()
	  + use new Maxcommandlength to split a request into multiple
	    subrequests then aggregate results before passing them
	    back to the caller
	- fetch_hash(): added checks for failed IMAP commands
	- parse_headers()
	  + properly check if fetch fails
	  + handle cases where $header and/or $field are not defined
	- size():
	  + return undef if LastError is set
	  + fix case where SIZE is not found and return undef as expected

version 3.16: Mon Apr  6 12:03:41 CEST 2009


	- set LastError when the imap_command receives an unexpected 'BYE' answer.
	  rt.cpan.org#44762 [Phil Lobbes]

	- handle SIGPIPE cleanly.
	  rt.cpan.org#43414 [Phil Lobbes]

	- improve handling of quotes in folder names
	  rt.cpan.org#43445 [Phil Lobbes]

	- do not use $socket->eof(), because IO::Socket::SSL does not support it.
	  rt.cpan.org#43415 [Phil Lobbes]

	- remove excessive reconfiguration of fastio in _read_line()
	  rt.cpan.org#43413 [Phil Lobbes]


	- remove experied docs about automatically created calls, which
	  do not exist since 3.00

	- remove verbose explanation about reporting bugs.

version 3.15: Fri Mar 20 13:20:39 CET 2009


	- manual-page was using POD syntax incorrectly, which caused many
	  broken links on search.cpan.org
	  rt.cpan.org #44212 [R Hubbell]

version 3.14: Mon Feb 16 14:18:09 CET 2009


	- isparent() when list() returns nothing.
	  rt.cpan.org#42932 [Phil Lobbes]

	- Quote more characters in Massage(): add  CTL, [, ], % and *
	  rt.cpan.org#42932 [Phil Lobbes]

	- message_string() will only complain about a difference between
	  reported message size and actually received size; it will not
	  try to correct it anymore.
	  rt.cpan.org#42987 [Phil Lobbes]

	- No error when empty text in append_string()
	  rt.cpan.org#42987 [Phil Lobbes]

	- login() should not try authenticate() if auth is empty or undef
	  rt.cpan.org#43277 [Phil Lobbes]

version 3.13: Thu Jan 15 10:29:04 CET 2009


	- "othermessage" in bodystructure parser should expect an MD5,
	  not bodyparams.  Fix and test(!) by [Michael Stok]


	- minor simplifications in code of run() and _imap_command()

	- get_bodystructure trace message fix [Michael Stok]

	- add Domain option for NTLM authentication.

version 3.12: Mon Nov 24 15:34:58 CET 2008


	- major performance improvement in append_message(), avoiding
	  reading the whole file in memory as the docs promised but the
	  code didn't do.  [David Podolsky]

version 3.11: Wed Oct  8 10:57:31 CEST 2008


	- some SSL connections process more bytes then needed, which
	  made the select() timeout.  Nice fix by [David Sansome]


	- improved example imap_to_mbox by [Ralph Sobek]

version 3.10: Sun Aug 24 21:26:27 CEST 2008


	- INET socket scope error, introduced by 3.09
	  rt.cpan.org#38689 [Matt Moen]

version 3.09: Fri Aug 22 16:38:25 CEST 2008


	- return status of append_message reversed.
	  rt.cpan.org#36726 [Jakob Hirsch]

	- no line-breaks in base64 encoded strings when logging-in
	  rt.cpan.org#36879 [David Jonas]

	- fix MD5 authentication.
	  rt.cpan.org#38654 [Thomas Jarosch]


	- extensions and clean-ups in examples/imap_to_mbox.pl by
	  [Ralph Sobek]

	- an absolute path as Server setting will open a local ::UNIX
	  socket, not an ::INET
	  rt.cpan.org#38655 [Thomas Jarosch]

version 3.08: Tue Jun  3 09:36:24 CEST 2008


	- message_to_file used wrong command.
	  rt.cpan.org#36184 [Parse Int]

	- oops, distribution released with OODoc/oodist, not make dist.
	  [Randy Harmon]

	- fix parsing of body-structure information for multi-parts.
	  rt.cpan.org#36279 [Doug Claar]


	- Updated README and TODO (Was 'Todo')

version 3.07: Mon Apr 28 09:17:30 CEST 2008


	- expunge with no folder specified produced "use of undef"
	  error.  Fixed by [André Warnier]

	- additional arguments for create [Michael Bacon]

	- accepts LIST answer with multiple lines [Michael Bacon]

	- ::BodyStructure::_address() should be _addresses()
	  Fixed by rt.cpan.org#35471 [Brian Kelly]

version 3.06: Mon Apr 14 23:44:03 CEST 2008


	- expunge without argument must use selected folder. [John W]

	- expunge with folder does not select it. [John W]

	- the documentation still spoke about "autogenerated methods",
	  but they were removed with 2.99 [John W]

	- append_string needs LF -> CRLF translations, for some
	  servers. rt.cpan.org #35031 [Jonathan Kamens]


	- added ::setquota(), thanks to [Jappe Reuling]

version 3.05: Wed Feb 20 08:59:37 CET 2008


	- match ENVELOPE and BODYSTRUCTURE more strict in the
	  grammar, to avoid confusion. [Zach Levow]

	- get_envelope and get_bodystructure failed for servers which
          did not return the whole answer in one piece. [Zach Levow]

	- do not produce parser errors when get_envelope does not
	  return an envelope. [Zach Levow]

	- PLAIN login response possibly solely a '+' [Zach] and [Nick]

version 3.04: Fri Jan 25 09:25:51 CET 2008


	- read_header fix for UID on Windows Server 2003.
          rt.cpan.org#32398 [Michiel Stelman]


	- doc update on authentication, by [Thomas Jarosch]

version 3.03: Wed Jan  9 22:11:36 CET 2008


	- LIST (f.i. used by folders()) did not return anything when the
	  passed argument had a trailing separator. [Gunther Heintze]

	- Rfc2060_datetime() must include a zone.
          rt.cpan.org#31971 [David Golden]

	- folders() uses LIST, and then calls a STATUS on each of the
	  names found.	This is superfluous, and will cause problems when
	  the STATUS fails... for instance because of ACL limitations
	  on the sub-folder.
	  rt.cpan.org#31962 [Thomas Jarosch]

	- fixed a zillion of problems in the BodyStructure parser.  The
	  original author did not understand parsing, nor Perl.

	- part numbering wrong when nested messages contained multiparts


	- implementation of DIGEST-MD5 authentication [Thomas Jarosch]

	- removed call for status() in Massage(), which hopefully speeds-up
	  things without destroying anything.  It removed a possible deep
	  recursion, which no-one reported (so should be ok to remove it)

	- simplified folders() algorithm.

	- merged folder commands, like subscribe into one.

	- added unsubscribe()
	  rt.cpan.org#31268 [G Miller]

	- lazy-load Digest::HMAC_MD5

version 3.02: Wed Dec  5 21:33:17 CET 2007


	- Another attempt to get get FETCH UID right.  Patch by [David Golden]

version 3.01: Wed Dec  5 09:55:43 CET 2007


	- removed version number from ::BodyStructure


	- quote password at login.
	  rt.cpan.org#31035 [Andy Harriston]

	- empty return of flags command should be empty list, not undef.
	  rt.cpan.org#31195 [David Golden]

	- UID command does not work with folder management commands
	  rt.cpan.org#31182 [Robbert Norris]

	- _read_line simplifications avoids timeouts.
	  rt.cpan.org#31221 [Robbert Norris]

	- FETCH did not detect the UID of a message anymore.
	  [David Golden]


	- proxyauth for SUN/iPlanet/NetScape IMAP servers.
	  patch by rt.cpan.org#31152 [Robbert Norris]

	- use grep in stead of map in one occasion in MessageSet.pm
	  [Yves Orton]

version 3.00: Wed Nov 28 09:56:54 CET 2007


	- "${peek}[]" should be "$peek\[]" for perl 5.6.1
	   rt.cpan.org#30900 [Gerald Richter]

version 2.99_07: Wed Nov 14 09:54:46 CET 2007


	- forgot to update the translate grammar.

version 2.99_06: Mon Nov 12 23:21:58 CET 2007


	- body structure can have any number of optional parameters.
	  Patch by [Gerald Richter].

	- get_bodystructure did not take the output correctly [Gerald Richter]

	- parser of body-structure did not handle optional body parameters
	  Patch by [Gerald Richter], rt.cpan.org#4479 [Geoffrey D. Bennet]

version 2.99_05: Mon Nov 12 00:17:42 CET 2007


	- pod error in MessageSet.pm

	- folders() without argument failed. [Gerald Richter]


	- better use of format syntax in date formatting.

	- Rfc2060_datetime also contains the time.

	- append_file() now has options to pass flags and time of file
	  in one go. [Thomas Jarosch]

version 2.99_04: Sat Nov 10 20:55:18 CET 2007


	- Simplified initiation of IMAP object with own Socket with a new
	  option: RawSocket [Flavio Poletti]


	- fixed read_line [Flavio Poletti]

	- fixed test-run in t/basic.t [Flavio Poletti]

version 2.99_03: Thu Nov  1 12:36:44 CET 2007


	- Remove note about optional Parse::RecDescent by Makefile.PL;
	  it is not optional anymore


	- When syswrite() returns 0, that might be caused by an error
	  as well.  Take the timeout/maxtemperrors track.
	  rt.cpan.org#4701 [C Meyer]

	- add NTLM support for logging-in, cleanly intergrated. Requires
	  the user to install Authen::NTLM.

version 2.99_02: Fri Oct 26 11:47:35 CEST 2007

	The whole Mail::IMAPClient was rewritten, hopefully without
	breaking the interface.  Nearly no line was untouched.

	The following things happened:
	 - use warnings, use strict everywhere
	 - removed many lines which were commented out, over the years
	 - $self->_debug if $self->Debug checked debug flag twice
	 - $self->LogError calls where quite inconsequent wrt $@ and carp
	 - consequent layout, changed sporadic tabs in blanks
	 - consequent calling convensions
	 - \0x0d\0x0a is always \r\n
	 - zillions of minor syntactical improvements
	 - a few major algorithmic rewrites to simplify the code, still
	   many oppotunities for improvements.
	 - expanded "smart" accessor methods, search abbreviations,
	   and autoloaded methods into separate subs.  In total much
	   shorter, and certainly better understandable!
	 - fixed many potential bugs.
	 - labeled some weird things with #????
	Over 1000 lines (30%!) and 25kB smaller in size
	Needs to be tested!!!! Volunteers?


	- Exchange 2007 only works with new parameter: IgnoreSizeErrors
	  rt.cpan.org#28933 [Dregan], #5297 [Kevin P. Fleming]

	- Passed socket did not get selected.
 	  debian bug #401144, rt.cpan.org# [Alexander Zanger],
	  #8480 [Karl Gaissmaier], #8481 [Karl Gaissmaier],
	  #7298 [Herbert Engelmann]

	- Seperator not correctly extracted from list command.
	  rt.cpan.org#9236 [Eugene Koontz], #4662 [Rasjid]

	- migrate() Massage'd foldername twice
	  rt.cpan.org#20703 [Peter J. Holzer]

	- migrate() could loop because error in regexp.
	  rt.cpan.org#20703 [Peter J. Holzer]

	- migrate() append_string result not tested.
	  rt.cpan.org#8577 [guest]

	- Failing fetch() returned undef, not empty list.
          rt.cpan.org#18361 [Robert Terzi]

	- Fix "use of uninitialised" warning when expunge is called
	  rt.cpan.org#15002 [Matt Jackson]

	- Fix count subfolders in is_parent, regexp did not take care
	  of regex special characters in foldername and seperator.
          rt.cpan.org#12883 [Mike Porter]

	- In fetch_hash(), the capturing of UID was too complicated
          (and simply wrong)
	  rt.cpan.org#9341 [Gilles Lamiral]

	- overload in MessageSet treated the 3rd arg (reverse) as

	- do not send the password on a different line as the username
	  in LOGIN.  Suggested by many people, amongst them
	  rt.cpan.org#4449 [Lars Uffmann]

	- select() with $timeout==0 (no timeout) returns immediately.
	  Should be 'undef' as 4th select parameter.
	  rt.cpan.org#5962 [Colin Robertson] and [Jules Agee]

	- examine() remembers Massage()d folder name, not the unescaped
	  version. rt.cpan.org#7859 [guest]


	- PREAUTH support by rt.cpan.org#17693 [Danny Siu]

	- Option "SupportedFlags", useful when the source supports
	  different flags than the peer in migrate().
	  Requested by rt.cpan.org#12961 [Don Christensen]

	- Fast_io did not clear $@ on unimportant errors.
	  rt.cpan.org#9835 [guest] and #11220 [Brian Helterline]

	- Digest::HMAC_MD5 and MIME::Base64 are now prerequisits.
	  rt.cpan.org#6391 [David Greaves]

	- PLAIN (SASL) authentication added, option Proxy
	  rt.cpan.org#5706 [Carl Provencher]

	- removed Bodystructure.grammar and IMAPClient.cleanup from dist.

	- reworked Bodystructure and MessageSet as well.

	- EnableServerResponseInLiteral now autodetect (hence ignored)

version 2.99_01:

	After 4 years of silence, Mark Overmeer took maintenance.  David
	Kernen could not be reached.  Please let him contact the new

	A considerable clean-up took place, fixing bug and adapting the
	distribution to current best practices.

	- use "prompt" in Makefile.PL, to please CPAN-testers

	- removed old Parse::RecDescent grammars

	- include Artistic and Copying (GPL) into COPYRIGHT file

	- remove INSTALL_perl5.80

	- removed all the seperate Makefile.PLs and test directories

	- removed the hard-copy of all involved RFCs: there are better
	  sources for those.

	- converted tests to use "Test::More"

	- Authmechanism eq 'LOGIN' understood.

	- test for CRAM-MD5 removed, because conflicts with test params
	  from Makefile.PL

	- test for fast-io removed, it is Perl core functionality

	- require IO::Socket::INET 1.26 to avoid Port number work-around.

	- Parse::RecDescent is required, and the grammars are pre-parsed
	  in the distribution.  This makes the whole installation process
	  a lot easier.

	- Update Todo, and many other texts.

	- added pod tester in t/pod.t

	- cleaned-up the rt.cpan.org bug-list from spam.  The next
	  release will contain fixes for the real reports.

Changes in version 2.2.9
Fixed problem in migrate that caused problems in versions of perl earlier
than 5.6.  Thanks go to Steven Roberts for reporting the problem and
identifying its cause.

Fixed problem in the make process that caused tests for BodyStructure
subclass to fail if the grammer had been compiled under a different
version of Parse::RecDescent. This problem was detected by the dedicated
people at testers@cpan.org.

Fixed a compatibility problem using Parse::RecDescent version 1.94.
This caused BodyStructure and Thread to fail for 5.8.x users. A number of
people reported this bug to CPAN but it took me a while to realize what
was going on. Really it took me a while to realize my Parse::RecDescent
was out of date. ;-) Now this module is delivered with two versions of
each of the affected grammars and Makefile.PL determines which version
to use.  Upgrading to Parse::RecDescent 1.94 will require you to re-run
Makefile.PL and reinstall Mail::IMAPClient.

Changes in version 2.2.8
Change the login method so that it always send password as a literal
to get around problem 2544 reported by Phil Tracy which caused
passwords containing asterisks to fail on some systems (but not any of
mine...). Good catch, Phil.

Added a new example that demonstrates the use of imtest (a utility
that comes with Cyrus IMAP) and Mail::IMAPClient together. The
example uses imtest to do secure authentication and then "passes" the
connection over to Mail::IMAPClient (but imtest is still brokering
the encryption/decryption). This example comes from an idea of
Tara L. Andrews', whose brainstorm it was to use imtest to broker
secure connections.  (But I still want to get encryption working with
Mail::IMAPClient some day!)

Fixed an error in which a "+" was used as a conncatenation error instead
of a ".".  Thanks to Andrew Bramble for reporting this, even though he
mistakenly identified it as a "typo". It is not a typo; a plus sign is the
correct concatenation operator, as any decent Java book will tell you ;-)

Fixed an error in the login method when the password contains a special
character (such as an asterisk.) Thanks to Phil Tracey for reporting
this bug.

Fixed some bugs in _send_line (the "O" side of the I/O engine) that were
reported by Danny Smith.

Fixed a bug in the migrate method in the optimization code (which 
gets called when socket writes are delayed due to a slow or busy target 
host, aka EAGAIN errors). Thanks to Pedro Carvalho for identifying 
this bug and its cause.

Fixed a bug in migrate that caused migration of unread messages to fail. 
This was due to the way Mail::IMAPClient's migrate method would try to send 
an empty list of flags to the target server in the APPEND. Thanks to 
Stephen Fralich at Syracuse University and for reporting this bug. 

Fixed another bug in the migrate method that caused flags to get lost. Thanks
go to Jean-Michel Besnard for reporting this.

Fixed a bug in migrate that caused
Fixed a bug in get_envelope that caused it to fail under certain conditions.
Thanks go to Bob Brown for reporting this bug.

Changes in version 2.2.7

Added some new parameters to support alternate authentication mechanisms:


Mail::IMAPClient has supported cram-md5 authentication "out of the box"
as of 2.2.6 (courtesy of Ville Skyttä). I also have digest-md5 working
in my lab with quality of protection levels "auth" and "integrity", but
not "confidentiality". I'm hoping to get the confidentiality part working
soon but so far have only managed to authenticate, send an encrypted command, 
and receive and decrypt the response. This may sound like enough but I can't 
seem to send a second command or receive a second response;-( In any event 
2.2.8 will support at least qop=auth and qop=auth-int but maybe not 

Fixed a bug reported by Adrian that caused get_bodystructure to
fail if the server returned a bodystructure with an embedded 
literal. Also fixed the same bug in get_envelope, so I guess now
everyone knows that get_envelope was just a tinkered-with copy of

Fixed two related bugs in Parser.pm that caused
get_bodystructure and get_envelope to fail if the
UID nnnnn part of a fetch response follows all the
other stuff.  Thanks to Raphaël Langella for reporting this bug.

Enhanced several methods to use MessageSets when the
Ranges parameter is true. There are still more methods that
need to be retrofitted to take advantage of the Range method
(and its underlying MessageSet object). In the meantime, if you 
need to get the functionality of the shorter message ranges provided
by the Range method from a method that does not honor the Ranges
parameter, then you should a) create a message set by passing the 
messages to the Range method and then pass the scalar as a string
to the method you want to use. For example, if you want to move 
a whole lot of messages to Trash, do something like this:
>my $range = $imap->Range(scalar($imap->search("SentBefore", "01-Jan-2000"))); 
This will cause the range object to stringify out to what looks like
a non-reference scalar before the move method gets the argument. If you
omit the quotes around "$range" then this won't work.

Fixed a bug in the list method that caused LIST "" "" to fail miserably.
Thanks to John W Sopko Jr. for reporting this bug.

Fixed a bug in the test suite that caused the cram-md5 tests to fail
if you are not running the extended tests. (Introduced in 2.2.6)

Fixed a bug that affected users on platforms that do not support 
fcntl (i.e. NT).  Thanks to Raphaël Langella for reporting this bug.

Changes in version 2.2.6

Fixed a bug in the migrate method that caused the internaldate
of migrated messages to sometimes be wrong. Credit goes to Jen Wu 
for identifying both bug and fix. 

Added a new method, "get_header", to provide a short-cut for a common
use of parse_headers. Added two other methods, "subject" and "date",
to provide shortcuts to get_header. 

Changed the Mail::IMAPClient::MessageSet module to override array 
dereferencing. (See below.)

Changed fetch and search methods to use the Range method (and thus the
Mail::IMAPClient::MessageSet module) for messages. The fetch method will
use MessageSet objects all the time, but the search method will only 
return MessageSet objects if you specify "Ranges => 1" (with Ranges being 
a new parameter).  The default will be "Ranges => 0" (which preserves 
the old behavior) but this default will go away in some future release.
There should be no need to override the fetch method's new behavior, since
it will be transparent to you unless you tend to fetch a lot of messages
at once, in which case your fetches may be faster and perhaps less likely
to fail due to the request exceeding your server's line limit. If you set
the Ranges parameter to true, then you still should not see a difference,
because a) when fetch is called in a list context then you will not get
a MessageSet object, you'll get the same list as always, and b) the 
MessageSet objects now override array de-referencing operations, so if you
treat the returned MessageSet object as if it were an array then the object
will humour you and act like a reference to an array of messages sequence
numbers or message uids.

Also changed the flags method to use the Range method. This should also
be transparent since the methods arguments and return values do not change.

Added built-in support for CRAM-MD5 authentication. This authentication 
method will in this release be used only when requested. In future releases
the default authentication will probably be the strongest authentication 
supported "out of the box" that is available on your server. Since CRAM-MD5
is the only authentication other than plain text that is currently supported
"out of the box", it will be the default authentication mechanism for any 
server that supports it. See the pod for the Authmechanism and Authcallback
parameters (which were also added in this release) and the doc for the 
authenticate method (which has been around a while). Many thanks to Ville Skyttä
for providing the code that makes up the heart of this new support, as well
as to Gisle Aas for the Digest::HMAC_MD5 and MIME::Base64.

Made minor tweaks to the documentation. Again. (Will it ever be 100% right?)

Changes in version 2.2.5
Added the Range method to convert a bunch of message UID's or sequence numbers
into compact ranges. Also added a supporting class for the returned range 
objects with overloaded operators that support stringifying, adding to, and
deleting from a range object's message set (Mail::IMAPClient::MessageSet). 
I also wrote documentation for same, so check it out. In future releases, 
I will probably enhance the base module to use MessageSet objects when 
feasible (i.e. whenever I know that the argument in question should in fact 
be a message specification). But I'll let you find all the bugs in the 
MessageSet module first ;-) Thanks goes to Stefan Schmidt, who is the first
to report using a server that restricted the size of a client request to 
something smaller than what Mail::IMAPClient was generating for him. 
(Originally the Range method was just supposed condense a message set into
the shortest possible RFC2060-compliant string, but then I got all happy and
started adding features. You know how it is...)

Changes in version 2.2.4
Fixed a bug in the done method (new in 2.2.3). 

Added tests for idle and done. (That's how I found the bug in the done method, above.)

Fixed minor bugs in test suite. (The test suite worked but wasn't always using the options
I wanted tested. <sigh>)

Changes in version 2.2.3

NOTE: This version was distributed to beta testers only.

Fixed the "Changes in version 2.2.2" section so that it correctly specifies
version 2.2.2 (instead of being yet another 2.2.1 section).

Fixed a bug in the migrate method that affected folders with spaces in their 

Fixed a bug in the Massage method that affected folders with braces ({}) in 
their names.

Added a new class method, "Quote", that will quote your arguments for you. (So you
no longer have to worry so much about quoting your quotes.

Added optimizations to the migrate method and to the core I/O engine inspired
by Jules Agee. (Actually they were not so much inspired by him as they were
lifted right out of a patch he had out on sourceForge.net. I had to refit them
for this version, and reformat his comments so they could fit in my window. Thanks
Jules, wherever you are.)

Added the fetch_hash method, which will fetch an entire folder's contents into a 
hash indexed by message UID (or message sequence number if that's all you've got).

Added a new example to the examples subdirectory, and corrected some minor bugs
in existing examples.

Added the idle and done methods, which together implement the IMAP IDLE extension
(RFC2177), at John Rudd's suggestion.

Changes in version 2.2.2
Fixed a bug in Massage method (generally only used by other IMAPClient methods)
that broke folder names with parens. 

Updated bug reporting procedures. Also added a section in the documentation 
for REPORTING THINGS THAT ARE NOT BUGS. Bug tracking is now done via 
rt.cpan.org, which I stumbled upon quite by accident and with which I am 
really pleased. A lot of credit goes to _somebody_ for putting this 
out on CPAN. Unfortunately as of this writing I don't whom.

Fixed a bug in the documentation regarding the logoff method, which is never
implicitly invoked anymore; I gave up on that because the DESTROY method would
sometimes be called after the Socket handle was already destroyed. (This is
especially likely at program exit, when everything still in scope goes out of
scope at the same time.) You should always log off explicitly if you want to 
be a well behaviod IMAP client.

Changes in version 2.2.1
Updated append_string to wrap the date argument in double quotes if the argument was
provided without quotes. Thanks to Grant Waldram for pointing out that some IMAP 
servers require this behavior.

Added a new method, selectable, which returns a true value if a folder is selectable.

Documented in this Changes file a change that was actually made for 2.2.0, in which 
newlines are chomped off of $@ (but not LastError).

Added pointers in the documentation to point to Mark Bush's Authen::NTLM module. This 
module will allow you to use NTML authentication with Mail::IMAPClient connections.
Also changed the authenticate method so that it will work with Authen::NTML without
the update mentioned in NTLM::Authen's README.

Added a second example on using the new migrate method,
migrate_mail2.pl. This example demonstrates more advanced techniques
then the first, such as using the separator method to massage folder
names and stuff like that.

Added support for the IMAP THREAD extension. Added
Mail::IMAPClient::Thread.pm to support this. (This pm file is generated
during make from Thread/Thread.grammar.) This new function should be
considered experimental. Note also that this extension has nothing to do
with threaded perl or anything like that. This is still on the TODO list.

Updated the search, sort, and thread methods to set $@ to "" before
attempting their respective operations so that text in $@ won't be left
over from some other error and therefore always indicative of an error
in search, sort, or thread, respectively.

Made many many tweaks to the documentation, including adding more examples
(albeit simple ones) and fixing some errors.

Changes in version 2.2.0
Fixed some tests so that they are less likely to give false negatives. For
example, test 41 would fail if the test account happened to have an
empty inbox.

Made improvements to Mail::IMAPClient::BodyStructure and renamed
Mail::IMAPClient::Parse to Mail::IMAPClient::BodyStructure::Parse. (This
should be transparent to apps since the ...Parse helper module is
used by BodyStructure.pm only.) I also resumed my earlier practice of
using ...Parse.pm from within BodyStructure.pm to avoid the overhead of
compiling the grammar every time you use BodyStructure.pm. (Parse.pm is
just the output from saving the compiled Parse::RecDescent grammar.) In a
related change, I've moved the grammar into its own file (Parse.grammar)
and taught Makefile.PL how to write a Makefile that converts the .grammar
file into a .pm file. This work includes a number of fixes to how a body
structure gets parsed and the parts list returned by the parts method,
among other things. I was able to successfully parse every bodystructure
I could get my hands on, and that's a lot.

Also added a bunch of new methods to Mail::IMAPClient::BodyStructure
and its child classes.  The child classes don't even have files of their
own yet; they still live with their parent class! Notable amoung these
changes is support for the FETCH ENVELOPE IMAP command (which was easy
to build in once the BODYSTRUCTURE stuff was working) and some helper
modules to get at the envelope info (as well as envelope information
for MESSAGE/RFC822 attachments from the BODYSTRUCTURE output).  Have a
look at the documentation for Mail::IMAPClient::BodyStructure for more

Fixed a bug in the folders method regarding quotes and folders with
spaces in the names. The bug must have been around for a while but
rarely manifested itself because of the way methods that take folder
name arguments always try to get the quoting right anyway but it was
still there. Noticing it was the hard part (none of you guys reported
it to me!).

Fixed a bug reported by Jeremy Hinton regarding how the search method
handles dates. It was screwing it all up but it should be much better now.

Added the get_envelope method which is like the get_bodystructure method
except for in ways in which it's different.

Added the messages method (a suggestion from Danny Carroll), which is
functionally equivalent to $imap->search("ALL") but easier to type.

Added new arguments to the bodypart_string method so that you can get
just a part of a part (or a part of a subpart for that matter...) I did
this so I could verify BodyStructure's parts method by fetching the first
few bytes of a part (just to prove that the part has a valid part number).

Added new tests to test the migrate function and to do more thorough
testing of the BodyStructure stuff. Also added a test to make sure that
searches that come up empty handed return an undef instead of an empty
array (reference), regardless of context. Which reminds me...

Fixed a bug in which searches that don't find any hits would return a
reference to an empty array instead of undef when called in a scalar
context. This bug sounds awfully familiar, which is why I added the test
mentioned above...

Changes in version 2.1.5
Fixed the migrate method so now it not only works, but also works
as originally planned (i.e. without requiring source messages to
be read entirely into memory).  If the message is smaller than
the value in the Buffer parameter (default is 4096) then a normal
$imap2->append($folder,$imap1->message_string) is done. However, if
the message is over the buffer size then it is retrieved and written a
bufferful at a time until the whole message has been read and sent. (The
receiving server still expects the entire message at once, but it
will have to wait because the message is being read from the source in
smaller chunks and then written to the destination a chunk at a time.)
This needs extensive testing before I'd be willing to trust it (or at
least extensive logging so you know when something has gone terribly
wrong) and I consider this method to be in BETA in this release. (Numerous
people wrote complaining that migrate didn't work, and some even included
patches to make it work, but the real bug in the last release wasn't
that migrate was broken but that I had inadvertently included the pod for
the method which I knew perfectly well was not ready to be released. My
apologies to anyone who was affected by this.) The migrate method does
seem to work okay on iPlanet (i.e. Netscape) Messenger Server 4.x. Please
let me know if you have any issues on this or any other platform.

Added a new example, migrate_mbox.pl, which will demonstrate the migrate method.

Fixed a bug that will cause Mail::IMAPClient's message reading methods to misbehave if 
the last line of the email message starts with a number followed by a space and either 
"OK", "NO", or "BAD".  This bug was originally introduced in 1.04 as a fix for another 
bug, but since the fix supports noncompliant behavior I'm disabling this behavior by 
default.  If your IMAP clients start hanging every time you try to read literal text 
(i.e. a message's test, or a folder name with spaces or funky characters) then you 
may want to turn this on with the EnableServerResponseInLiteral parameter. Thanks go 
to Manpreet Singh for reporting this bug.

Fixed a bug in imap_to_mbox.pl that has been there since 2.0.0 (when the Uid 
parameter started defaulting to "True"). Thanks to Christoph Viethen for reporting 
the bug and suggesting the fix. BUT NOTE THIS: I often don't test the example programs,
so you should think of them as examples and not free production programs. Eventually 
I would like to add tests to my test suite (either the 'make test' test suite that you
run or my own more extensive test suite) but it's not a super high priority right now.

Significant improvements to the whole Mail::IMAPClient::BodyStructure module 
were contributed by Pedro Melo Cunha. It's really much better now.

Bullet-proofing added to some private methods. (Private meaning they are undocumented
and not part of the module's API. This is perl not java.)

Fix applied to unset_flag to support user-defined flags (thanks to E.Priogov 
for submitting the bug report and patch).

Changes in version 2.1.4
Added Paul Warren's bugfix to the sort method.

Added Mike Halderman's bugfix for the get_bodystructure method.

Fixed a localization problem reported by Ivo Panecek. Because of this fix,
the Errno.pm file is now a prerequisite to this module. This way I can just
test to see if the error is an "EAGAIN" error (as defined in sys/errno.h and thus
Errno.pm) instead of awkwardly checking the string value of $!. 

I also renamed the MaxTempErrors parameter to Maxtemperrors in response the same 
bug report.  Added a "MaxTempErrors" accessor method that will set and return 
Maxtemperrors for backwards compatibility. Also, the number of temporary errors 
gets reset after each successful I/O, so that the socket i/o operation fails only if
you if your temporary I/O errors happen more than "Maxtemperrors" times in a row.
The old behavior was to continue incrementing the count of temporary errors until
either the entire message was written or until a total of Maxtemperrors had occurred,
regardless of how many intervening successful syswrites occurred. This was a bug, but 
Ivo politely suggested the new behavior as an enhancement. ;-) Also, you can now
specify "UNLIMITED" as the Maxtemperrors, in which case these errors will be ignored.
And the default for Maxtemperrors is now 100, but I'm open to any feedback you may
have in this regard. 

I also fixed the operator precedence problem that was reported by many folks in that
very same part of the code. (As you may have guessed, that code was new in the last

One of the people who reported the precedence problem was Jules Agee, who also submitted
a patch that may in the end provide an optimal solution to handling EAGAIN errors. 
Unfortunately I have not had time to retrofit his patch into the current version of the
module. But if I can manage to do this soon and it tests well I'll include it in the next
release, in which case the Maxtemperrors parameter will be of interest only to historians.

I also received a patch from John Ello that adds support for Netscape's proprietary 
PROXYAUTH IMAP client command. I haven't included that support in this release because 
you can already use the proxyauth method. It's one of those famous "default" methods 
that, despite their fame and my documentation, nobody seems to know about. But you 
can always say "$imap->proxyauth($uid)", for example, providing that $imap and $uid 
are already what they're supposed to be. (I've been doing this myself for years.)

However, John's patch does provide a cleaner interface (it remembers who you are as 
well as who you were, for example) so I may include it later as part of a separate 
module that extends Mail::IMAPClient. This would also give me an excuse for providing 
the framework for plugging in Administrative methods that are proprietary to other imap 
servers, so if you have a technique for acquiring administrative access to your users' 
mailboxes (besides proxyauth) please let me know what it is. Perhaps we'll get something cool out of it, like a document on how to write administrative scripts for various 
platforms and a suite of supporting methods for each. 

Changes in version 2.1.3
Added the new method append_string. It works similarly to append but will allow extra 
arguments to supply the flags and internal date of the appended message. See the pod 
for more details.

(Thanks to Federico Edelman Anaya for suggesting this fix.)

Fixed a bug in the AUTOLOAD subroutine that caused "myrights" (and possibly other 
non-existant methods) to fail. Thanks go to Larry Rosenbaum for reporting the bug 
and identifying the fix.

Added the new method Escaped_results, which preprocesses results so that data 
containing certain special characters are returned quoted with special characters 
(like quotes!) escaped. (I needed this for the bodystructure stuff, below.)

NEW! Added support for parsing bodystructures (as provided in the server response to 
FETCH BODYSTRUCTURE). This support requires Parse::RecDescent and is implemented via two 
new modules, Mail::IMAPClient::BodyStructure and Mail::IMAPClient::Parse. Note that 
the latter module is used by the former; your programs need not and should not use it 
directly so don't. Also, these modules are ALPHA and EXPERIMENTAL so no screaming when 
they don't work. (Polite bug reports will of course be gratefully accepted.) Many 
thanks to Damian Conway, the author of Parse::RecDescent, without which this feature 
would not have been possible (or at least not very likely).

Enhanced support for DOS systems (and DOS's offspring, such as windows) by removing 
the "\c\n"s and replacing them with "\x0d\x0a". Thanks go to Marcio Marchini for his 
help with this effort.

Fixed the list of symbols imported along with Fcntl.pm. (Paul Linder asked me to put 
this in the last release but I forgot.)

Changes in version 2.1.2

Fixed a bug in the is_parent method which made it inaccurate on some servers.

Added new method "sort", which implements the SORT extenstion and which was contributed 
by Josh Rotenberg. The SORT extension is documented at 
http://search.ietf.org/internet-drafts/draft-ietf-imapext-sort-06.txt. A copy of the 
draft is also included with the Mail::IMAPClient distribution, which means I also:

Added draft-ietf-imapext-sort-06.txt to the docs subdirectory of the distribution.

Fixed a bug in the folders method and the subscribed method (same bug, appeared twice) 
which broke these methods under some conditions. Thanks again Josh Rotenberg for supplying the fix.

Fixed bugs in getacl and listacl. Changed the interface for getacl significantly; 
existing scripts using getacl will not behave the same way. But then on the other hand, 
getacl was never documented before, so how could you be using it? 

Implemented improvements to reduce memory usage by up to 30%. Thanks go Paul Linder, 
who developed the memory usage patch after a considerable amount of analysis. The 
improvements include the use of 'use constant', so your perl needs to support that 
pragma in order to use Mail::IMAPClient.

Added a new parameter, MaxTempErrors, which allows the programmer to control the number 
of consecutive "Resource Temporarily Unavailable" errors that can occur before a write 
to the server will fail. Also changed the behavior of the client when one of these 
errors occurs. Previously, Mail::IMAPClient waited .25 seconds (a quarter of one 
second) before retrying the read operation. Now it will wait (.25 * the number of 
consecutive temporary errors) seconds before retrying the read. 

Documented the "Buffer" parameter, which has been secretly available for some time. I 
just forgot to document it. It sets the size of the read buffer when Fast_io is turned 
on. (NOTE: As of version 2.1.5 it also controls the size of the buffer used by the 
migrate method.) 

Updated the Todo file. It was nice to see that a number of lines in the "Todo" file were now deletable. It was depressing to see that a number of original lines need to stay 
in there. 

Changes in version 2.1.1
Added the "mark", "unmark", and imap4rev1 methods.

Updated the documentation to include the new methods and to document "create", "store", 
and "delete".

Updated "message_string" to be smart about whether you're using IMAP4 or IMAP4REV1.

Updated "message_to_file" to be smart about whether you're using IMAP4 or IMAP4REV1.

Added several bug fixes to authenticate method. Many thanks to Daniel Wright who 
reported these bugs and provided the information necessary to fix them.

Changes in version 2.1.0

Fixed a serious bug introduced in 2.0.9 when appending large messages.

Made minor changes to improve the cyrus_expunge.pl example script.

Made the set_flags routine RFC2060-compliant. Previously it prepended flag names with 
backslashes, even if the flags were not reserved flags. This broke support for 
user-defined flags, which I didn't realize was supposed to even be there until Scott 
Renner clued me in. (Thanks, Scott.)

Promoted the release level to "1". 

Added a new 'internaldate' method. (Thanks to the folks at jwm3.org for donating the 

Added a new example, cyrus_expire.pl.

Changes in version 2.0.8/2.0.9
Made minor changes to the tests in t/basic.t so that folders are explicitly closed 
before they are deleted. (Don't worry, only folders created by the tests are 
deleted. :-) Thanks go to Alan Young for reporting that some servers require this.

Changed the routine that massages folder names into IMAP-compliant strings so that 
single-quotes in a name do not force the folder to go through as "LITERAL" strings 
(as defined in RFC2060). This shouldn't cause a problem for anybody (and in fact 
should make life easier for some folks) but if you do have any trouble with 
single-quotes in folder names PLEASE LET ME KNOW ASAP!! 

Divided the sending of literal strings into two I/O operations (as required by RFC2060).
This should correct problems with sending literals to some servers that will not read 
any data sent before they reply with the "+ go ahead" message. (Thanks go to Keith Clay,
who reported seeing this problem with the M-Store IMAP server.)

Changed the "create" method so that it will autoquote the first argument to create 
rather than the last. Normally the first argument is the last, but Cyrus users can 
specify an optional 2nd argument, except when using pre-2.0.8 versions of 
Mail::IMAPClient ;-) Thank you Chris Stratford for reporting this bug and 
identifying its cause.

Fixed a bug in body_string when the message is empty. (Thanks go to Vladimir Jebelev for 
finding this bug and providing the fix.)

Added a new example to the examples subdirectory. cyrus_expunge.pl is a script you 
can use (after making minor tweaks) to periodically expunge your server's mail store. 

Changes in version 2.0.7
Fixed a bug in message_count. Thanks go to Alistair Adams for reporting this bug.

Fixed a bug in folders that caused some foldernames to not be reported in the 
returned array.

Changes in version 2.0.6

Applied patches from Phil Lobbe to tighten up sysreads and 'writes and to correct a
bug in the I/O engine.

Changes in version 2.0.5

Fixed bug in parse_headers so that RFC822 headers now match the pattern /(\S*):\s*/
instead of /(\S*): /. Thanks go to Paul Warren for reporting this bug and providing the

Added more robust error checking to prevent infinite loops during read attempts and
fixed bugs in parse_headers. Thanks go to Phil Lobbes, who provided several useful 
patches and who performed valuable pre-release testing.

Changes in version 2.0.4

Fixed bug in parse_headers when connected to an Exchange server with UID=>1. (Kudos to 
Wilber Pol for that fix.)

Fixed bugs in parse_headers and tightened reliability of I/O engine by implementing 
many improvements suggested by Phil Lobbes, who also provided code for same.

Added bugfix that under certain conditions caused server responses to be "repeated"
when fast_io is turned on. Thanks to Jason Hellman for providing bug report and
diagnostic data to fix this.

Added a "LastIMAPCommand" method, which returns the last IMAP client command that 
was sent to the server. 

Removed the "=begin debugging" paragraph that somehow got included in CPAN's 
html pages (even though it shouldn't have). 

Began a process of redesigning the documentation. I would like to be able to present
a more formal syntax for the various methods and hope to have that ready for the next

Tested successfully against Cyrus v 2.0.7.

Tested unsuccessfully against mdaemon. This appears to be due to mdaemon's 
noncompliance with rfc2060 so future support for mdaemon should not be expected 
any time soon.  ;-(

Changes in version 2.0.3

Did major rewrite of message_string method, which should now be both cleaner 
and more reliable.

Fixed bug in move method that caused some folders to be incorrectly quoted. 
Thanks go to Felix Finch for reporting this bug. Also, at his suggestion I 
added information to move documentation explaining the need to expunge.

Made many fixes and tweaks to pod text.

Added a new method, Rfc2060_date, which takes times in the "seconds since 1/1/1970" 
format and returns a string in RFC2060's "dd-Mon-yyyy" format (which is the format 
you need to use in IMAP SEARCH commands).

Changes in version 2.0.2
Fixed bug that caused a compile error on some earlier versions of perl5. 

Noticed that some older versions of perl give spurious "Ambiguous use" warnings 
here and there, mostly because I'm not quoting the name of the "History" member 
of the underlying Mail::IMAPClient hash. These warnings will go away when you upgrade 
perl. (I may fix them later, or maybe not. Depends on if I have time.) 

Added new parameter (and eponymous method) Peek, along with new tests for 'make test' 
for same. See the pod for further info.

Added some error checking to avoid trying to read or write with an 
unconnected IMAPClient object.

Made bug fixes to parse_headers and flags.

Added missing documentation for the exciting new message_to_file method (oops). 
Also cleaned up a few typos in the pod while I happened to be there. (I'm sure 
there are still plenty left.)

Fixed bugs in append and append_file. (Thanks to Mauro Bartolomeoli and to the people 
at jwm3.org for reporting these bugs.) 

Made changes to call to syswrite to guarantee delivery of entire message. (Only affects 
appends of very large messages.)

Added the 'close' method to the list of lower-case-is-okay methods (see the section 
under version 2.0.0 on "NEW ERROR MESSAGES").

Changes in version 2.0.1
Several bug fixes related to the flags method and to spurious warning messages 
when run with warnings turned on.

A new method, message_to_file, writes message text directly into a file. This 
bypasses saving the text in the history buffer and the overhead that entails, which 
could be especially important when processing big ass messages. Of course the bad news 
is that now you'll have to write all that shtuff out to a filehandle, but maybe you 
wanted to do that anyway. Anyhow, between append_file and message_to_file, both 
of which take filehandle arguments, there should be a way to "short circuit" the 
copying of mail between two imap sessions. I just haven't got it completely figured 
out yet how it would work. Got any ideas? Anyhow, this method is currently considered 

A couple of new tests have been added to go along with our new little method.

I've added a whole bunch more IMAP-related rfc's to the docs/ subdirectory. Trust me, 
you are going to need them.

Changes in version 2.0.0
This version includes a major rewrite of the I/O engine. It's now cleaner and more 
reliable. Also, output processing is less likely to match patterns that look like 
server output but are really, say, message text contained in a literal or something 
like that. Also, various problems with blank lines at the ends of messages either 
magically appearing or disappearing should now go away. Basically, it's much better 
is what I'm trying to say.

The Uid parameter now defaults to true. This should be transparent to existing scripts 
(except for those scripts that produce embarrassing results because someone forgot to 
specify Uid=>1, in which case they'll magically start behaving somehow).

The namespace method has been added, thus implementing RFC2342. If you have any scripts 
that rely on the old, "default method" style of namespace implementation then you should 
rename those method calls to be mixed case (thus forcing the AUTOLOADed default method). 

Mail::IMAPClient now issues a lot more warning messages when run in warn mode 
(i.e. $^W is true). Of particular interest are methods implemented via the "default 
method" AUTOLOAD hack. They will generate a warning telling you to use mixed- or 
upper-case method names (but only if warnings are turned on, say with the -w switch 
or $^W++ or something).  The exceptions are certain unimplemented yet quite popular 
methods that, if ever explicitly implemented, will behave the same way as they do via 
the default method. (Or at least they will remain downwardly compatible. I may add 
bells and whistles by not by default.) Those methods are listed in the pod and right 
here:  store, copy, subscribe, close, create, delete and expunge.

Changed the version numbering scheme to match perl's (as of perl v5.6.0).

Added a few new tests to the test suite. (Still need more, though.) Also changed fast_io 
and uidplus test suites so that they just "do" the basic tests but with different 
options set (i.e. Fast_io and Uid, respectively).

- The expunge method now optionally accepts the name of the folder to be expunged. It's 
also been documented, even though it technically doesn't exist. (That won't stop it from 
working, though.) Since expunge deletes messages that you thought were already deleted, 
it's only appropriate to use a method that you thought existed but really doesn't, don't 
you think? And if you're wondering how I managed to change the behavior of a method that 
doesn't exist, well, I don't want to talk about it.

- Speaking of methods that don't exist (also known as methods implemented via "the 
default method"), effective with this release there are a number of unimplemented 
methods that are guaranteed to always exhibit their current behavior. In other words, 
even if I do eventually implement these methods explicitly, they will continue to 
accept the same arguments and return the same results that they do now via the default 
method. (Why I would even bother to do that is specifically not addressed in this 
document.) Currently this means that these methods will not trigger warnings when 
called via all-lowercase letters (see "NEW ERROR MESSAGES", above). In the future I 
hope that it will also mean that these non-existant but functioning methods will also 
be documented in the pod.

- Fixed a bug in the flags method introduced in 1.19.  (Thanks to the people at jwm3.org 
for reporting this!)

Changes in version 1.19
Fixed a bug in which the Folder parameter returned quoted folder names, which sometimes 
caused other methods to requote the folders an extra time. (The IMAP protocol is real 
picky about that.) Thanks go to Felix Finch for both reporting the bug and identifying 
the fix.

Siggy Thorarinsson contributed the new "unseen_count" method and suggested a new 
"peek mode" parameter.  I have not yet gotten around to implementing the new parameter 
but have included the unseen_count method, since a) he was kind enough to write it, and 
b) it tests well. 

In the meantime, you cannot tell methods like "parse_headers" and "message_string" and 
so forth whether or not you want them to mark messages as "\Seen". So, to make life 
easier for you in particular I added a bunch of new methods: set_flag, unset_flag, 
see, and deny_seeing. The latter two are derivitives of the former two, respectively, 
which should make this sentence almost as difficult to parse as an IMAP conversation.

Fixed bug in which "BAD" "OK" or "NO" lines prefixed by an asterisk (*) instead of the 
tag are not handled correctly. This is especially likely when LOGIN to a UW IMAP server 
fails. Thanks go to Phil Lobbes for squashing this bug.

Fixed bug in logout that caused the socket handle to linger. Credit goes to 
Jean-Philippe Bouchard for reporting this bug and for identifying the fix.

Fixed bug in uidvalidity method where folder has special characters in it.

Made several bug fixes to the example script examples/find_dup_msgs.pl. Thanks to Steve 
Mayer for identifying these bugs.

Changed Fast_io to automatically turn itself off if running on a platform that does 
not provide the necessary fcntl macros (I won't mention any names, but it's initials 
are "NT"). This will occur silently unless warnings are turned on or unless the Debug 
parameter is set to true. Previously scripts running on this platform had to turn off 
fast_io by hand, which is lame. (Thank you Kevin Cutts for reporting this problem.)

Updated logic that X's out login credentials when printing debug output so that funky 
characters in "User" or "Password" parameters won't break the regexp. (Kevin Cutts found 
this one, too.)

Tinkered with the Strip_cr method so it can accept multiple arguments OR an array 
reference as an argument. See the updated pod for more info.

Fixed a typo in the documentation in the section describing the fetch method. There 
has been an entire paragraph missing from this section for who knows how long. Thanks 
to Adam Wells, who reported this documentation error.

Fixed bug in seen, recent, and unseen methods that caused them to return empty arrays 
erroneously under certain conditions.

Changes in version 1.18
Timeouts during read operations now work correctly.

Fixed several bugs in the I/O engine. This should correct various problems with Fast_io 
turned on (which is now the default).

Reworked message_string and body_string methods to avoid bugs when Uid set to true.

Changes in version 1.17

Added support for the Oracle IMAP4r1 server.

Tinkered with the DESTROY method so that it does a local($@) before doing its evals. 
This will perserve the value of $@ when the "new" method fails during a login but the 
DESTROY's "logout" succeeds. The module was setting the $@ variable, but on some 
versions of perl the DESTROY method would clobber $@ before anything useful could be 
done with it! Thanks to Kimmo Hovi for reporting this problem, which was harder to 
debug than you might think.

Changes in version 1.16

IMPORTANT:  Made Fast_IO the default. You must specify Fast_io => 0 in your new method 
call or invoke the Fast_io method (and supply 0 as an arg) to get the old behavior. 
(This should be transparent to most users, but as always your mileage may vary.)

Reduced the number of debug msgs printed in the _read_line internal method and added a 
debug msg to report perl and Mail::IMAPClient versions.

The message_count method will now return the number of messages in the currently select 
folder if no folder argument is supplied. 

The message_string method now does an IMAP FETCH RFC822 (instead of a 
FETCH RFC822.HEADERS and a FETCH RFC822.TEXT), which should eliminate missing blank 
lines at the ends of some messages on some IMAP server platforms. It also returns undef 
if for some reason the underlying FETCH fails (i.e. there is no folder selected), 
thanks to a suggestion by Pankaj Garg. It has also been slightly re-worked to support 
the changes in the I/O engine from version 1.14.

Re-worked the body_string method to support the I/O engine changes from v1.14.

Fixed a bug in parse_headers when used with multiple headers and the Uid parameter set 
to a true value.

Documented in this file a fix for a bug in the flags method with the Uid parameter 
turned on. (Belated thanks to Michael Lieberman for reporting this bug.)

Changes in version 1.15
Fixes the test suite, which in v1.14 had an "exit" stmt that caused early termination 
of the tests.  (I had put that "exit" in there on purpose, and left it in there by 

Changes in version 1.14
Fixed a bug in the _readline subroutine (part of the I/O engine) that was caused by my 
less-than-perfect interpretation of RFC2060. This fix will allow the Mail::IMAPClient 
module to function correctly with servers that imbed literal datatypes in the middle 
of response lines (rather than just at the end of them). Thanks to Pankaj Garg for 
reporting this problem and providing the debugging output necessary to correct it.

Fixed a bug in parse_headers that was introduced with the fix to the I/O engine 
described above.

Changes in version 1.13
Changed the parse_headers method so that it uses BODY.PEEK instead of BODY. This 
prevents the parse_headers method from implicitly setting the "\Seen" flag for messages 
that have not been otherwise read. This change could produce an incompatibility in 
scripts that relied on the parse_headers previous behavior. 

Fixed a bug in the flags method with the Uid parameter turned on.  (Thanks to Michael 
Lieberman for reporting this bug.)

Changes in version 1.12
Fixed a bug in the folders method when called first with a second arg and then without 
a second arg.

Tested sucessfully with perl-5.6.0. 

Added a section to the pod documentation on how to report bugs. I've had to ask for 
output from scripts with "Debug => 1" so many times that I eventually decided to 
include the procedure for documenting bugs in the distribution. (Duh! It only took me 
11 releases to come up with that brainstorm.) Often following the procedures to obtain 
the documentation is enough; once people see what's going on (by turning on Debug =>1) 
they no longer want to report a bug. 

Did I mention it's a good idea to turn on debugging when trying to figure out why a 
script isn't working? (It is.)

In order to make the Debug parameter friendlier, it now prints to STDERR by default. 
You can override this by supplying the spanking brand new Debug_fh parameter, which 
if supplied had better well point to a filehandle (either by glob or by reference), 
and by 'filehandle' I mean something besides STDIN!

Debugging mode will now also X-out the login credentials used to login. This will make 
it easier to share your debugging output.

Added documentation for the State parameter, which must be set manually by programmers 
who are not using Mail::IMAPClient's connect and/or login methods but who are instead 
making their own connections and then using the Socket parameter to turn their 
connections into IMAP clients.

Fixed bug in parse_headers with Uid turned on.

Fixed bug in parse_headers when using the argument "ALL".

Changes in version 1.11
Added new example script, copy_folder.pl, to demonstrate one way to copy entire 
folders between imap accounts (which may or may not be on the same server). This 
example is right next to all the others, in the examples/ subdirectory of the 

Changed error handling slightly. $@ now contains pretty much the same stuff as what 
gets returned by LastError, even when LastError won't work (i.e. when an implicit 
connect or login fails and so no object reference is returned by new). You can thank 
John Milton for the friendly nagging that got me to do this.

Added new test suite for the fast_io engine. This should make it easier to determine 
whether or not the fast_io engine will work on your platform. 

Implemented a work-around to allow the Port parameter to default despite a known bug in 
IO::Socket::INET version 1.25 (distributed with perl 5.6.0).

Fixed a bug in the message_string method in which the resulting text string for some 
mime messages to be incompatible with append.

Fixed a bug in the Fast_io i/o engine that could cause hangs during an append operation.

Changed a number of regular expressions to accept mixed-case "Ok", "No" or "Bad" 
responses from the server and to do multi-line matching.

Fixed a bug in the append method that was causing extra carriage returns to appear in 
messages whose lines were already terminated with the CR-LF sequence. Thanks to Heather 
Adkins for reporting this bug.

Enhanced the parse_headers routine so that it is less sensitive to variations of 
case in message headers. Now, the case of the returned key matches the case of the 
field as specified in the parse_headers method's arguments, regardless of its case 
in the message being parsed. (You can thank Heather Atkins for this suggestion as 
well.) See below for more changes to parse_headers in this release.

Improved the append method so that it has better error handling and error recovery. 
Thanks to Mark Keisler for pointing out some bugs in the error handling code in 
this method.

Added the append_file method, which is like the append method but it works on files 
instead of strings.  The file provided to append must contain an RFC822-formatted 
message. Use of the append_file method avoids having to stuff huge messages into 
variables before appending them. Thanks to jwmIII (http://jwm3.org) for suggesting 
this method.

Changed the flags method and the parse_headers method so that a reference to an array 
of message sequence numbers (or message UIDS if the Uid parameter is turned on) can 
optionally be passed instead of a single message sequence number (or UID). Use of this 
enhancement will change your return values so be sure to read the pod.  Thanks to 
Adrian Smith (adrian.smith@ucpag.com) for delivering this enhancement.

Fixed a bug in "message_string" that caused the blank lines between headers and body 
to fall out of the string.

Tinkered with the undocumented _send_line method to permit an optional argument 
to suppress the automatic insertion of <CR><LF> at the end of strings being sent. 
(NOTE: I'm telling you this because I'm a nice guy. This doesn't mean that _send_line 
is now a programming interface.)

Changes in version 1.10

Added two new methods, lsub and subscribed. lsub replaces the behavior of the default 
method and should be downwardly compatible. The subscribed method works like the 
folders method but the results include only subscribed folders. Thanks to Alexei 
Kharchenko for providing the code for lsub (which is the foundation upon which 
'subscribed' was built).

Changes in version 1.09

Changed login method so that values for the User parameter that do not start and end 
with quotes will be quoted when sent to the server. This is to support user id's 
with embedded spaces, which are legal on some platforms. 

Changed name of test input file created by perl Makefile.PL and used by 'make test' 
from .test to test.txt to support weird, offbeat OS platforms that cannot handle 
filenames beginning with a dot.

Fixed bugs in seen, unseen, and recent methods.  (These are almost the same method 
anyway; they are dynamically created at compile time from the same code, with 
variable substitution filling in the places where "seen", "unseen", or "recent" 
belong.) The bug caused these methods to return the transaction number of the 
search as if it were the last message sequence number (or message uid) in
the result set.

Added the 'since' method, which accepts a date in either standard perl format (seconds 
since 1/1/1970, or as output by time and as accepted by localtime) or in the date_text 
format as defined in RFC2060 (dd-Mon-yyyy, where Mon is the English-language 
three-letter abbreviation for the month). It searches for items in the currently 
selected folder for messages sent since the day whose date is provided as an argument.

Added 'sentsince', 'senton', 'sentbefore', 'on', and 'before' methods which are 
totally 100% just like the 'since' method, except that they run different searches. 
(Did I mention that it's useful to have RFC2060 handy when writing IMAP clients?)

Added two new methods, run and tag_and_run, to allow IMAP client programmers finer 
control over the IMAP conversation. These methods allow the programmer to compose 
the entire IMAP command string and pass it as-is to the IMAP server. The difference 
between these two methods is that the run method requires that the string include 
the tag while the tag_and_run method requires that it does not.

To a similar end, the pre-existing Socket parameter and eponymous accessor method 
has been documented to allow direct access to the IMAP socket handle and to allow 
the socket handle to be replaced with some other file handle, presumably one derived 
from a more interesting technology (such as SSL).

Fixed a bug that caused blank lines to be removed from 'literal' output (as defined 
in RFC2060) when fast_io was not used. This bug was especially likely to show up in 
routines that fetched a message's body text.  The fact that this bug did not occur 
in the newer fast_io code may indicate that I've learned something, but on the other 
hand we shouldn't jump to rash conclusions.

I've run benchmarks on the fast_io code to determine whether or not it is faster and, 
if so, under what circumstances. It appears that the fast_io code is quite faster, 
except when reading large 'literal' strings (i.e. message bodies), in which case it 
appears to take the same amount of time as the older i/o code but at the cost of 
more cpu cycles (which means it may actually be slower on cpu-constrained systems). 
The reason for this is that reads of literal strings are by their nature already 
optimized, but without the overhead of fcntl calls. So if you expect to be doing 
lots of message text (or multipart message body parts) fetching you should not use 
fast_io, but in pretty much any other case you should go ahead and use it. In any 
event, a number of people have tested fast_io so I no longer consider it 
experimental, unless you're running perl on NT or CP/M or something funky like that, 
in which case let me know how you make out! 

Changes in version 1.08

Maintenance release 1.08a fixes a bug in the folders method when supplying the 
optional argument (see "Enhanced folders method..." below) with some IMAP servers.

Added option to build_ldif.pl (in the examples subdirectory) to allow new options and 
to better handle quoted comments in e-mail addresses. Thanks to Jeffrey Fiedl, 
whose book _Mastering Regular Expressions_ (O'Reilly) helped me to figure out a 
good way to do this.

Fixed documentation error that failed to mention constraints on when the append 
method will return the uid of the appended message. (This feature only works with 
servers that have the UIDPLUS capability.)

Added/improved documentation somewhat.

The copy method now returns a comma-separated list of uids if successful and if the 
IMAP server supports UIDPLUS extentions. The move method now works similarly.

Added new method uidnext, which accepts the name of a folder as an argument and returns 
the next available message UID for that folder.

The exists and append methods now will handle unquoted foldernames with embedded 
spaces or quotes or whatever. Including quotes as part of the argument string is no 
longer required but is still supported for backwards compatibility reasons. In other 
words, $imap->exists(q("Some Folder")) is now no longer necessary (but will still work). $imap->exists(some folder) is good enough.

Mail::IMAPClient has been tested successfully on Mirapoint 2.0.2. (Thanks to Jim 

I've now installed the UW imapd IMAP4rev1 v12.264 on one of my machines so I'm better 
able to certify that platform. All the tests in 'make test' work there (or are at least 
gently skipped). 

Fixed bug in getacl in which folder names were quoted twice. (Thanks to Albert Chin for 
squashing this bug.) Similar bugs existed in the other ACL methods and were similarly 

Fixed a bug in message_uid that basically caused it to not work. Muchos gracias to 
Luvox (aka fluvoxamine hydrochloride) for providing me with just the help I needed to 
discover and fix this bug. 

Enhanced folders method to allow an argument. If an argument is supplied, then 
the folders method will restrict its results to subfolders of the supplied argument 
(which should be the name of a parent folder, IMHO). This is implemented by supplying 
arguments to the LIST IMAP Client command so we are optimizing network I/O at the 
expense of possible server incompatibilities.  If you find server incompatibilities 
with this then please let me know, and in the meantime you can always 
grep(/^parent/,$imap->folders) or something. Or re-implement the folders 
method yourself.

Changes in version 1.07
Added a new parameter, Fast_io, which, if set to a true value, will attempt to 
implement a faster I/O engine. USE THIS AT YOUR OWN RISK. It is alpha code. I don't 
even know yet if it even helps.

Added support for spaces in folder names for the autoloaded subscribe method.

Added new methods setacl, getacl, deleteacl, and listrights. These methods are not yet 
fully tested and should be considered beta for this release.

Enhanced support for the myrights method (which is implemented via the default method).

Fixed bug in append method that caused it to hang if server replied to original APPEND 
with a NO (because, say, the mailbox's quota has been exceeded).

Removed the autodiscovery of the folder hierarchy from the login method. This will 
speed up logging in but may delay certain other methods later (but see the next item, 

Updated the exists method to issue a "STATUS" IMAP Client command, rather than depend 
on the folder hierarchy being discovered via 'LIST "" "*"'. Apparently this speeds 
things up a lot for some configurations, although the difference will be negligable to 

Updated Makefile.PL to support the PREFIX=~/ directive. Thanks to Henry C. Barta 
(hbarta@wwa.com) for this fix.

Added the Timeout parameter and eponymous accessor method, which, if set to a true 
value, causes reads to time out after the number of seconds specified in the Timeout 
parameter. The value can be in fractions of a second. This has not been fully tested 
though, so use of this parameter is strictly "Beta".

Enhanced support for the UID IMAP client command. Setting the new Uid parameter to a 
true value will now cause the object to treat all message numbers as message UID 
numbers rather than message sequence numbers. Setting the Uid parameter to a false 
value will turn off this behavior again. 

Updated test suite to handle servers that cannot do UIDPLUS and to add tests for 
the Uid parameter.

Incorporated bug fixes for recent_count and message_count in which some servers are 
sticking in extra \r's, and updated DESTROY to remove spurious warning messages under 
some versions of perl (thanks to Scott Wilson for catching and killing these bugs).

Changes in version 1.06
Changed folders method so that it correctly handles mail folders whose names start and 
end with quotes.

Changed append method so that it returns the uid of the newly appended message if
successful. Since the uid is a "true" value this should not affect the behavior of
existing scripts, although it may enhance the behavior of new scripts ;-)

Fixed bug in parse_headers that could cause script to die if there were no headers of
the type requested and if there was a space on the blank line returned from FETCH.
(Some blank lines are blanker than others...)

Added the "flags" method, which returns an array (or array reference if called in scalar
context) containing the flags that have been set for the message whose sequence number 
has been provided as the argument to the method. 

Added the "message_string" method, which accepts a message sequence number as an 
argument and returns the contents of the message (including RFC822 headers) as a 
single string.

Added the "body_string" method, which accepts a message sequence number as an argument 
and returns the contents of the message (not including RFC822 headers) as a single 

Changes in version 1.05

Patched the 'make test' basic test to work correctly on systems that do not
support double quotes in folder names. Thanks to Rex Walters for this fix.

Added a new example script, build_dist.pl, that rumages through a folder 
(specified on the command line) and collects the "From:" address, and then 
appends a message to that folder with all those addresses in both the To: field 
and the text, to facilitate cuting and pasting (or dragging and dropping) 
into address books and so forth. (Note that the message doesn't actually get 
sent to all those people; it just kind of looks that way.)

Also added another example, build_ldif.pl, that is similar to build_dist.pl 
except that instead of listing addresses in the message text, it creates a 
MIME attachment and attaches a text file in LDIF format, which can then be
imported into any address book that supports LDIF as an import file format.
This example requires the MIME::Lite module. MIME::Lite was written by Eryq 
(okay, Erik Dorfman is his legal name), and is totally available on CPAN. 

This distribution has now been tested on Mirapoint Message Server Appliances
(versions 1.6.1 and 1.7.1). Many thanks to Rex Walters for certifying this
platform and for providing a test account for future releases.

Changes in version 1.04

Fixed situation in which servers that include the "<tag> <COMMAND> OK\r\n" line
as part of a literal (i.e. text delivered via {<length>}\r\n<length> bytes\r\n)
caused the module to hang. This situation is pretty rare; I've only run across
one server that does it. I'm sure it's a bug; I'm not sure whose. ;-}
Many thanks to Thomas Stromberg for 1) pointing out this bug and 2) providing 
me with facilities to find and fix it!

Fixed potential bug in I/O engine that could cause module to hang when reading 
a literal if the first read did not capture the entire literal.

Cleaned up some unnecessary runtime warnings when a script is executed with 
the -w switch.

Added new tests to 'make test'. I just can't keep my hands off it! ;-)

Enhanced the append method and several tests in 'make test' to be more widely
compatible. Successfully tested on UW-IMAP, Cyrus v1.5.19, Netscape Messenger 
4.1, and Netscape Messenger v3.6. If you know of others please add them to 
the list! 

Fixed a bug in the separator method (new in 1.03) that caused it to fail if
'inbox' was specified in lowercase characters as the method's argument. 

Added a new example, imap_to_mbox.pl, contributed by Thomas Stromberg. This 
example converts a user's IMAP folders on an IMAP server into mbox format. 

Changes in version 1.03
Reworked several methods to support double-quote characters within folder 
names. This was kind of hard. This has been successfully tested with create, 
delete, select, and folders, to name the ones that come to mind.

Reworked the undocumented method that reads the socket to accept and handle
more gracefully lines ending in {nnn}\r\n ( where nnn is a number of 
characters to read). This seems to be part of the IMAP protocol although I 
am at a total loss as to where it's explained, other than a brief description 
of a "literal's" bnf syntax, which hardly counts. 

Added separator object method, which returns the separator character in use 
by the current server. 

Added is_parent method, which returns 1, 0, or undef depending on whether a 
folder has children, has no children, or is not permitted to have children.

Added tests to 'make test' to test new function. Also changed 'make test' to 
support IMAP systems that allow folders to be created only in the user's INBOX 
(which is the exact opposite of what my IMAP server allows...oh, well).

Fixed a bug that caused search to return an array of one undef'ed element 
rather than undef if there were no hits.

Changes in version 1.02
Fixed bugs in search and folders methods. 

Fixed bug in new method that ignored Clear => 0 when specified as arguments to 

Changes in version 1.01
Fixed a bug in test.pl that caused tests to fail if the extended tests were not used.

Added method 'parse_headers' to parse the header fields of a message in the 
IMAP store into a perl data structure.

Changes in version 1.00
Made cosmetic changes to documentation.

Fixed a bug introduced into the 'folders' method in .99.

Changed 'new' method so that it returns undef if an implicit connection or 
login is attempted but fails. Previous releases returned a Mail::IMAPClient 
object that was not connected or not logged in, depending on what failed.

Changed installation script so that it reuses the parameter file for test.pl 
if it finds one. Installation can be run in the background if the test.txt file 
exists. Touching it is good enough to prevent prompts; having a correctly 
formatted version (as described in test_template.txt) is even better, as it will 
allow you to do a thorough 'make test'.

Changes in version .99 
Added the Rfc822_date class method to create RFC822-compliant date fields in 
messages being appended with the append method.

Added the recent, seen, and unseen methods to return an array of sequence 
numbers from a SEARCH RECENT, SEARCH SEEN, or SEARCH UNSEEN method call. 
These methods are shortcuts to $imap->search("RECENT"), etc. 

Added the recent_count method to return the number of RECENT messages in a 
folder. Contributed by Rob Deker.

Added 'use strict' compliance, courtesy of Mihai Ibanescu.

Fixed a bug in the search method that resulted in a list with one empty member 
being returned if a search had no hits. The search method now returns undef 
if there are no hits.

Added 'authenticate' method to provide very crude support for the IMAP 
AUTHENTICATE command. The previous release didn't support AUTHENTICATE at all, 
unless you used very low-level (and undocumented) methods. With the 
'authenticate' method, the programmer still has to figure out how to 
respond to the server's challenge. I hope to make it friendlier in the 
next release. Or maybe the one after that. This method is at least a start, 
albeit a pretty much untested one.

Added Rfc822_date class method to facilitate creation of "Date:" header 
field when creating text for the "append" method, although the method may 
come in handy whenever you're creating a Date: header, even if it's not 
in conjuction with an IMAP session.

Added more tests, which will optionally run at 'make test' time, provided all 
the necessary data (like username, hostname, password for testing an IMAP 
session) are available.

Changes in version 0.09  
Thu Aug 26 14:10:03 1999 - original version; created by h2xs 1.19

# $Id: Changes,v 20001010.18 2003/06/12 21:35:48 dkernen Exp $