1.0.39:
      - New configuration variable only_hosts allows
	you to limit check_mk to a subset of your
	hosts (for testing)
      - New configuration parameter mem_extended_perfdata
	sends more performance data on Linux (see 
	check manual for details)
      - many improvements of Multiadmin web pages: optionally 
	filter out services which are (not) currently in downtime
	(host or service itself), optionally (not) filter out summary
	hosts, show host status (down hosts), new action
	for removing all scheduled downtimes of a service.
	Search results will be refreshed every 90 seconds.
	Choose between two different sorting orders.
	Multadmin now also supports user authentication
      - New configuration option define_timeperiods, which
	allows to create Nagios timeperiod definitions.
	This also enables the Multiadmin tools to filter
	out services which are currently not in their
	notification interval.
      - NIC check for Linux (netctr.combined) now supports
	checking of error rates
      - fc_brocade_port: New possibility of monitoring
	CRC errors and C3 discards
      - Fixed bug: snmp_info_single was missing
        in precompiled host checks
	
1.0.38:
      - New: check_mk's multiadmin tool (Python based
	web page). It allows mass administration of
	services (enable/disable checks/notifications, 
	acknowledgements, downtimes). It does not need
	Nagios service- or host groups but works with
	a freeform search.
      - Remove duplicate <?php from the four new 
	PNP templates of 1.0.37.
      - Linux Agent: Kill hanging NFS with signal 9
	(signal 15 does not always help)
      - Some improvements in autodetection. Also make
	debug mode: ./autodetect.py: This helps to
	find problems in autodetection.
      - New configuration variables generate_hostconf and
	generate_dummy_commands, which allows to suppress
	generation of host definitions for Nagios, or 
	dummy commands, resp.
      - Now also SNMP based checks use cache files.
      - New major options --backup and --restore for
	intelligent backup and restore of configuration
	and runtime data
      - New variable simulation_mode allows you to dry
	run your Nagios with data from another installation.
      - Fixed inventory of Linux cpu.loads and cpu.threads
      - Fixed several examples in checks manpages
      - Fixed problems in install_nagios_on_lenny.sh
      - ./setup.sh now understands option --yes: This
        will not output anything except error messages
	and assumes 'yes' to all questions
      - Fix missing 'default.php' in templates for
	local
	
1.0.37:
      - IMPORTANT: Semantics of check "cpu.loads" has changed.
	Levels are now regarded as *per CPU*. That means, that
	if your warning level is at 4.0 on a 2 CPU machine, then 
	a level of 8.0 is applied.
      - On check_mk -v now also ouputs version of check_mk
      - logfile_patterns can now contain host specific entries.
	Please refer to updated online documentation for details.
      - Handling wrapping of performance counters. 32 and 64 bit
	counters should be autodetected and handled correctly.
	Counters wrapping over twice within one check cycle
	cannot be handled, though.
      - Fixed bug in diskstat: Throughput was computed twice
	too high, since /proc/diskstats counts in sectors (512 Bytes)
	not in KB
      - The new configuration variables bulkwalk_hosts and
	non_bulkwalk_hosts, that allow 	to specify, which hosts 
	support snmpbulkwalk (which is
	faster than snmpwalk) and which not. In previos versions,
	always bulk walk was used, but some devices do not support
	that.
      - New configuration variable non_aggregated_hosts allows
	to exclude hosts generally from service aggregation.
      - New SNMP based check for Rittal CMC TC 
	(ComputerMultiControl-TopConcept) Temperature sensors 
      - Fixed several problems in autodetection of setup
      - Fixed inventory check: exit code was always 0
	for newer Python versions.
      - Fixed optical problem in check manual pages with
	newer version of less.
      - New template check_mk-local.php that tries to
	find and include service name specific templates.
	If none is found, default.php will be used.
      - New PNP templates check_mk-kernel.php for major page
	faults, context switches and process creation
      - New PNP template for cpu.threads (Number of threads)
      - Check nfsmounts now detects stale NFS handles and
	triggers a warning state in that case

1.0.36:
      - New feature of Linux/UNIX Agent: "MRPE" allows
	you to call Nagios plugins by the agent. Please
	refer to online documentation for details.
      - Fix bug in logwatch.php: Logfiles names containing spaces
	now work.
      - Setup.sh now automatically creates cfg_dir if
	none found in nagios.cfg (which is the case for the
	default configuration of a self compiled Nagios)
      - Fix computation of CPU usage for VMS.
      - snmp_hosts now allows config-list syntax. If you do
	not define snmp_hosts at all, all hosts with tag
	'snmp' are considered to be SNMP hosts. That is 
	the new preferred way to do it. Please refer
	to the new online documentation.
      - snmp_communities now also allows config-list syntax
	and is compatible to datasource_programs. This allows
	to define different SNMP communities by making use
	of host tags.
      - Check ifoperstatus: Monitoring of unused ports is
	now controlled via ifoperstatus_monitor_unused.
      - Fix problem in Windows-Agent with cluster filesystems:
	temporarily non-present cluster-filesystems are ignored by
	the agent now.
      - Linux agent now supports /dev/cciss/d0d0... in section
	<<<diskstat>>>
      - host configuration for Nagios creates now a variable
	'name host_$HOSTNAME' for each host. This allows
	you to add custom Nagios settings to specific hosts
	in a quite general way.
      - hosts' parents can now be specified with the
	variable 'parents'. Please look at online documentation
	for details.
      - Summary hosts now automatically get their real host as a
	parent. This also holds for summary cluster hosts.
      - New option -X, --config-check that checks your configuration
	for invalid variables. You still can use your own temporary
	variables if you prefix them with an underscore.
	IMPORTANT: Please check your configuration files with
	this option. The check may become an implicit standard in
	future versions.
      - Fixed problem with inventory check on older Python 
	versions.
      - Updated install_nagios_on_lenny.sh to Nagios version
	3.2.0 and fixed several bugs.

1.0.35:
      - New option -R/--restart that does -S, -H and -C and
	also restarts Nagios, but before that does a Nagios
	config check. If that fails, everything is rolled
	back and Nagios keeps running with the old configuration.
      - PNP template for PING which combines RTA and LOSS into
	one graph.
      - Host check interval set to 1 in default templates.
      - New check for hanging NFS mounts (currently only
	on Linux)
      - Changed check_mk_templates.cfg for PING-only hosts:
	No performance data is processed for the PING-Check
	since the PING data is already processed via the
	host check (avoid duplicate RRDs)
      - Fix broken notes_url for logwatch: Value from setup.sh
	was ignored and always default value taken.
      - Renamed config variable mknagios_port to agent_port
	(please updated main.mk if you use that variable)
      - Renamed config variable mknagios_min_version to
	agent_min_version (update main.mk if used)
      - Renamed config variable mknagios_autochecksdir to 
	autochecksdir (update main.mk if used)
      - configuration directory for Linux/UNIX agents is
	now configurable (default is /etc/check_mk)
      - Add missing configuration variable to precompiled
	checks (fix problem when using clusters)
      - Improved multipath-check: Inventory now determines
	current number of paths. And check output is more
	verbose.
      - Mark config files as config files in RPM. RPM used
	to overwrite main.mk on update!
	
1.0.34:
      - Ship agents for AIX and SunOS/Solaris (beta versions).
      - setup script now autodetects paths and settings of your
	running Nagios
      - Debian package of check_mk itself is now natively build
	with paths matching the prepackaged Nagios on Debian 5.0
      - checks/df: Fix output of check: percentage shown in output
	did include reserved space for root where check logic did
	not. Also fix logic: account reserved space as used - not
	as avail.
      - checks/df: Exclude filesystems with size 0 from inventory.
      - Fix bug with host tags in clusters -> precompile did not
	work.
      - New feature "Inventory Check": Check for new services. Setting
	inventory_check_interval=120 in main.mk will check for new services
	every 2 hours on each host. Refer to online documentation
	for more details.
      - Fixed bug: When agent sends invalid information or check
	has bug, check_mk now handles this gracefully
      - Fixed bug in checks/diskstat and in Linux agent. Also
	IDE disks are found. The inventory does now work correctly
	if now disks are found.
      - Determine common group of Apache and Nagios at setup.
	Auto set new variable www_group which replaces logwatch_groupid.
	Fix bug: logwatch directories are now created with correct
	ownership when check_mk is called manually as root.
      - Default templates: notifications options for hosts and
	services now include also recovery, flapping and warning
	events.
      - Windows agent: changed computation of RAM and SWAP usage
	(now we assume that "totalPageFile" includes RAM *and*
	SWAP).
      - Fix problem with Nagios configuration files: remove
	characters Nagios considers as illegal from service
	descriptions.
      - Processing of performance data (check_icmp) for host
        checks and PING-only-services now set to 1 in default
	templates check_mk_templates.cfg.
      - New SNMP checks for querying FSC ServerView Agent: fsc_fans,
	fsc_temp and fsc_subsystems. Successfully tested with agents
	running	on Windows and Linux.
      - RPM packaged agent tested to be working on VMWare ESX 4.0 
	(simply install RPM package with rpm -i ... and open port 
	in firewall with "esxcfg-firewall -o 6556,tcp,in,check_mk")
      - Improve handling of cache files: inventory now uses cache
	files only if they are current and if the hosts are not
	explicitely specified.
	
1.0.33:
      - Made check_mk run on Python 2.3.4 (as used in CentOS 4.7
	und RedHat 4.7). 
      - New option -M that prints out manual pages of checks.
	Only a few check types are documented yet, but more will
	be following.
      - Package the empty directory /usr/lib/check_mk_agent/plugins
	and ../local into the RPM and DEB package of the agent
      - New feature: service_dependencies. check_mk lets you comfortably
	create Nagios servicedependency definitions for you and also
	supports them by executing the checks in an optimal order.
      - logwatch.php: New button for hiding the context messages.
	This is a global setting for all logfiles and its state is
	stored in a cookie.
	
1.0.32:
      - IMPORTANT: Configuration variable datasource_programs is now
        analogous to that of host_groups. That means: the order of
        program and hostlist must be swapped!
      - New option --fake-dns, useful for tests with non-existing
	hosts.
      - Massive speed improvement for -S, -H and -C
      - Fixed bug in inventory of clusters: Clustered services where
	silently dropped (since introduction of host tags). Fixed now.
      - Fixed minor bug in inventory: Suppress DNS lookup when using
	--no-tcp
      - Fixed bug in cluster handling: Missing function strip_tags()
	in check_mk_base.py was eliminated.
      - Changed semantics of host_groups, summary_host_groups,
	host_contactgroups, and summary_host_groups for clusters: 
	Now the cluster names will be relevant, not
	the names of the nodes. This allows the cluster hosts to
	have different host/contactgroups than the nodes. And it is more
	consistent with other parts of the configuration.
      - Fixed bug: datasource_programs on cluster nodes did not work
	when precompiling

1.0.31:
      - New option -D, --dump that dumps all configuration information
	about one, several or all hosts
	New config variables 'ignored_checktypes' and 'ignored_services',
        which allow to include certain checktypes in general or
        some services from some hosts from inventory
      - Config variable 'clustered_services' now has the same semantics
	as ignored_checktypes and allows to make it host dependent.
      - Allow magic tags PHYSICAL_HOSTS, CLUSTER_HOSTS and ALL_HOSTS at
	all places, where lists of hosts are expected (except checks).
	This fixes various problems that arise when using all_hosts at
	those places:
	  * all_hosts might by changed by another file in conf.d
	  * all_hosts does not contain the cluster hosts
      - Config file 'final.mk' is read after all other config files -
	if it exists. You can put debug code there that prints the
	contents of your variables.
      - Use colored output only, if stdout is a tty. If you have
	problems with colors, then you can pipe the output
	through cat or less
      - Fixed bug with host tags: didn't strip off tags when
	processing configuration lists (occurs when using
	custom host lists)
      - mk_logwatch is now aware of inodes of logfiles. This
	is important for fast rotating files: If the inode
	of a logfile changes between two checks mk_logwatch
	assumes that the complete content is new, even if
	the new file is longer than the old one.
      - check_mk makes sure that you do not have duplicate
	hosts in all_hosts or clusters.

1.0.30:
      - Windows agent now automatically monitors all existing
	event logs, not only "System" and "Application".

1.0.29:
      - Improved default Nagios configuration file:
	added some missing templates, enter correct URLs
	asked at setup time.
      - IMPORANT: If you do not use the new default 
	Nagios configuration file you need to rename
	the template for aggregated services (summary
	services) to check_mk_summarizes (old name
	was 'check_mk_passive-summary'). Aggregated
	services are *always* passive and do *never*
	have performance data.
      - Hopefully fixed CPU usage output on multi-CPU
	machines
      - Fixed Problem in Windows Agent: Eventlog monitoring
	does now also work, if first record has not number 1
	(relevant for larger/older eventlogs)
      - Fixed bug in administration.html: Filename for Nagios
	must be named check_mk.cfg and *not* main.mk. Nagios
	does not read files without the suffix .cfg. 
      - magic factor for df, that allows to automatgically 
        adapt levels for very big or very small filesystems.
      - new concept of host tags simplyfies configuration.
      - IMPORTANT: at all places in the configuration where
	lists of hosts are used those are not any longer
	interpreted as regular expressions. Hostnames
	must match exactly. Therefore the list [ "" ] does
	not any longer represent the list of all hosts.
	It is a bug now. Please write all_hosts instead
	of [ "" ]. The semantics for service expressions
	has not changed.
      - Fixed problem with logwatch.php: Begin with
	<?php, not with <?. This makes some older webservers
	happy.
      - Fixed problem in check ipmi: Handle corrupt output
	from agent
      - Cleaned up code, improved inline documentation
      - Fixed problem with vms_df: default_filesystem_levels,
	filesystem_levels and df magic number now are used
	for df, vms_df and df_netapp together. Works now also
	when precompiled.
	
1.0.28:
      - IMPORTANT: the config file has been renamed from
	check_mk.cfg to main.mk. This has been suggested
	by several of my customers in order to avoid 
	confusion with Nagios configuration files. In addition,
	all check_mk's configuration file have to end in
	'.mk'. This also holds for the autochecks. The 
	setup.sh script will automatically rename all relevant
	files. Users of RPM or DEB installations have to remove
	the files themselves - sorry.
      - Windows agent supports eventlogs. Current all Warning
        and Error messages from 'System' and 'Application' are
        being sent to check_mk. Events can be filtered on the
	Nagios host.
      - Fixed bug: direct RRD update didn't work. Should now.
      - Fixed permission problems when run as root.
      - Agent is expected to send its version in <<<check_mk>>>
	now (not any longer in <<<mknagios>>>
      - Fixed bug in Windows agent. Performance counters now output
	correct values
      - Change checks/winperf: Changed 'ops/sec' into MB/s.
	That measures read and write disk throughput
	(now warn/crit levels possible yet)
      - new SNMP check 'ifoperstatus' for checking link
        of network interfaces via SNMP standard MIB
      - translated setup script into english
      - fixed bug with missing directories in setup script
      - made setup script's output nicer, show version information
      - NEW: mk_logwatch - a new plugin for the linux/UNIX agent
	for watching logfiles
      - Better error handling with Nagios pipe
      - Better handling of global error: make check_mk return
	CRIT, when no data can retrieved at all.
      - Added missing template 'check_mk_pingonly' in sample
	Nagios config file (is needed for hosts without checks)
	
1.0.27:
      - Ship source code of windows agent
      - fix several typos
      - fix bug: option --list-hosts did not work
      - fix bug: precompile "-C" did not work because
	of missing extension .py
      - new option -U,--update: It combines -S, -H and
	-U and writes the Nagios configuration into a
	file (not to stdout).
      - ship templates for PNP4Nagios matching most check_mk-checks.
	Standard installation path is /usr/share/check_mk/pnp-templates
	
1.0.26:
      -	Changed License to GNU GPL Version 2
      - modules check_mk_admin and check_mk_base are both shipped
	uncompiled.
      - source code of windows agent togehter with Makefile shipped
	with normal distribution
      - checks/md now handles rare case where output of /proc/mdstat
	shows three lines per array

1.0.25:
      - setup skript remembers paths

1.0.24:
      - fixed bug with precompile: Version of Agent was always 0

1.0.23:
      - fixed bug: check_config_variables was missing in precompiled
	files
      - new logwatch agent in Python plus new logwatch-check that
	handles both the output from the old and the new agent

1.0.22:
      - Default timeout for TCP transfer increased from 3.0 to 60.0
      - Windows agent supports '<<<mem>>>' that is compatible with Linux
      - Windows agents performance counters output fixed
      - Windows agent can now be cross-compiled with mingw on Linux
      - New checktype winperf.cpuusage that retrieves the percentage
	of CPU usage from windows (still has to be tested on Multi-CPU
	machine)
      - Fixed bug: logwatch_dir and logwatch_groupid got lost when
	precompiling. 
      - arithmetic for CPU usage on VMS multi-CPU machines changed

1.0.21:
      - fixed bug in checks/df: filesystem levels did not work
	with precompiled checks

1.0.20:
      - new administration guide in doc/
      - fixed bug: option -v now works independent of order
      - fixed bug: in statgrab_net: variable was missing (affected -C)
      - fixed bug: added missing variables, imported re (affected -C)
      - check ipmi: new option ipmi_summarize: create only one check for all sensors
      - new pnp-template for ipmi summarized ambient temperature
 
1.0.19:
      - Monitoring of Windows Services
      - Fixed bug with check-specific default parameters
      - Monitoring of VMS (agent not included yet)
      - Retrieving of data via an external programm (e.g. SSH/RSH)
      - setup.sh does not overwrite check_mk.cfg but installs
	the new default file as check_mk.cfg-1.0.19
      - Put hosts into default hostgroup if none is configured
