2006-04-19  Berend de Boer  <berend@pobox.com>

	* Release 0.7.4 of emc.


2006-04-17  Berend de Boer  <berend@pobox.com>

	* Binc imap server included a flag that should not have been
	included in the flag list according to RFC 3501. The imap4
	response parser didn't like this. This flag is now accepted.


2006-02-17  Berend de Boer  <berend@pobox.com>

	* Fix precondition failure when parsing date past Unix time.


2005-06-21  Berend de Boer  <berend@pobox.com>

	* Sources adapted to change in string formatting classes in Gobo.


2004-10-22  Berend de Boer  <berend@pobox.com>

	* secure option in config file always handled as true (due to
	is_equal between UC_STRING and STRING, sigh).


2004-03-13  Berend de Boer  <berend@pobox.com>

	* Release 0.7.0 of emc.

	* Secure connection to imap (imaps) is now supported. You need to
	openssl program for this, which should be installed if you install
	openssl.

	* Configuration now allows to set port in emc.config

	* You can now override the spam probablity assigned to a token
	that is not in the token list.


2004-02-28  Berend de Boer  <berend@pobox.com>

	* Last word of subject and body wasn't tokenized.


2004-01-05  Berend de Boer  <berend@dellius.nederware.nl>

	* The code to tokenize the character set used in an email did
	nothing. The intention was to treat the character set as a special
	token that would get a high probability of being spam it it was in
	your spam corpus but not in your non-spam corpus.

	* emc.rnc and emc.xsd were not present in the binary releases.

	* emc_scan incorrectly reported that probabilities.txt had to be
	moved if the configuration file resided in the current directory.

	* When emc.config was not found, the error message was incorrectly
	saying that emc.conf was not found.


2003-12-31  Berend de Boer  <berend@dellius.nederware.nl>

	* Release 0.6.0 of emc.


2003-12-30  Berend de Boer  <berend@dellius.nederware.nl>

	* emc_classify now exits when it is passed an argument it does not
	recognize.

	* Added minimum deviation range to emc.config file. Words with a
	probability in this range are not used to determine the spaminess
	of an email.


2003-12-16  Berend de Boer  <berend@pobox.com>

	* If both a <move-nonspam-to> and <move-spam-to> mailbox were
	provided in the emc.config file, the debug version would fail with
	a precondition failure. Thanks to Doug Stevens for the report.


2003-12-08  Berend de Boer  <berend@pobox.com>

	* emc_scan prints proper login failure instead of just aborting.


2003-12-02  Berend de Boer  <berend@pobox.com>

	* Generates special tokens if header and body failed to parse
	correctly.


2003-09-20  Berend de Boer  <berend@pobox.com>

	* Prints percentage spam/nonspam statistics after classifying.

	* Failure to retrieve header resulted in planned exit. Now prints
	error and continues.


2003-08-22  Berend de Boer  <berend@pobox.com>

	* Release 0.5.2 of emc.


2003-08-21  Berend de Boer  <berend@pobox.com>

	* The probabilities.txt file is now searched in the directory
	where emc.config is found.

	* Configuration file emc.config is now searched first in the
	current directory, next in ~/.emc/ and than in
	/usr/local/etc/emc/.


2003-08-20  Berend de Boer  <berend@pobox.com>

	* Bug in my base64 decoder routines caused base 64 encoded
	messages to be skipped.

	* Windows binaries had a strange limit on the number of
	attachment: only 48 attachments were supported. This was due to
	the Borland C compiler I'll compile with the Microsoft C compiler
	from now on.


2003-08-12  Berend de Boer  <berend@pobox.com>

	* Release 0.5 of emc.


2003-06-28  Berend de Boer  <berend@pobox.com>

	* Added attribute "minimum-occurrences" to scan. Default is one,
	set to a higher value to only safe tokens that occur at least
	that many times. With value two you can get rid of all those
	tokens that just occur once.

	* emc_classify read only 256 character long lines from
	probabilities.txt. However, there was no check in emc_scan to
	write a word that was restricted to around 240 characters. Now
	emc_scan will write only words of up to 64 characters.
	This is done by limiting the tokenizer to not generate more than 64
	characters. So words longer than 64 characters will be equal if
	the first 64 characters are equal.


2003-06-07  Berend de Boer  <berend@pobox.com>

	* Release 0.4, first public release.
