initial "rules" for the new zabbix templates at https://zabbix.org/wiki/Docs/specs/ZBXNEXT-973
Just read the spec, great 
Can I suggest that templates should use macros for trigger thresholds that people may want to change for individual hosts (such as load average, or number of processes)
Yep, we are keeping that in mind for new templates.
New set of templates was integrated into trunk revision 26979, more to come.
Any chance they could be put somewhere so we can import them?
I have noticed three issues with the new templates.
(1) Template SNMP Disks lacks graph prototypes for discovered storages. Graph prototypes can be defined in the same way as for Mounted filesystem discovery in Template OS Linux.
<Sasha> RESOLVED
(2) Same as the above issue with Template SNMP Processors.
<Sasha> SNMP returns only one value for CPU Load - The average, over the last minute. There graph prototypes are not needed.
(3) The default graph width of 400 pixels is not enough for disk space usage graphs, because partition/filesystem/volume names can be quite long.
<Sasha> RESOLVED
Changing assignee.
(4) SNMP values sysName, sysDescr, sysContact, sysLocation as well as data from 'uname -a' could be used to populate corresponding host inventory fields by default.
<Sasha> RESOLVED
(5) Templates FreeBSD, OpenBSD and Mac OS X lack low level discovery rules for network interfaces.
<Sasha> FreeBSD, OpenBSD and Mac OS X does not support low level discovery of network interfaces
(6) There is a graph for Zabbix Server Performance in the Template Application Zabbix Server. This graph currently displays only one value - items processed per second. Previously there was Zabbix queue value on that graph too.
<Sasha> RESOLVED
(7) There are unsupported items in Template OS Windows:
system[procrunning]
perf_counter\\PhysicalDisk(_Total)\\Avg. Disk Read Queue Length
perf_counter\\PhysicalDisk(_Total)\\Avg. Disk Write Queue Length
perf_counter\\System\\File Read Bytes/sec
perf_counter\\System\\File Write Bytes/sec
perf_counter\\System\\threads
<Sasha> RESLOVED
Should "system[procrunning]" be changed to "proc.num[,,run]"?
<Sasha> RESOLVED item "system[procrunning]" has been removed
The PerfCounter keys depend on the operating system language. They should all be changed to the corresponding numbers:
perf_counter\\PhysicalDisk(_Total)\\Avg. Disk Read Queue Length --> perf_counter[\234\(_Total)\1402]
perf_counter\\PhysicalDisk(_Total)\\Avg. Disk Write Queue Length --> perf_counter[\234\(_Total)\1404]
perf_counter\\System\\File Read Bytes/sec --> perf_counter[\2\16]
perf_counter\\System\\File Write Bytes/sec --> perf_counter[\2\18]
perf_counter\\System\\threads --> perf_counter[\2\250]
proc.num item accepts only 2 parameters for Windows hosts. Example: proc.num[] - works, proc.num[,] - works, proc.num[,,] - ZBX_NOTSUPPORTED.
As far as there is currently no way to get the number of running processes under Windows, I think, that this item should be removed from Template OS Windows.
Thanks, Jens, numerical perf_counters work well.
(8) In Template OS Windows in graph prototype for disk usage items for total disk space and free disk space should be interchanged with eachother, otherwise disk usage graph is incorrect.
<Sasha> RESOLVED
Can new templates for items and formulas associate device class from MIB/OID snmpwalk for discovery policy to automatically create inherit groups of hosts, items, applications, and services? Perhaps this is already being integrated into advanced discovery?
There are no predefined graphs for memory usage in OS templates.
I made a template for APC Network Management Card 2.
How can I submit it?
please, create a new zbxnext and attach it there
you can also upload it in zabbix.org wiki and link to it from https://zabbix.org/wiki/Zabbix_Templates
In Template OS Linux the Incoming/Outgoing traffic items (net.if.in/out) have a user multiplier of 8. As I understand, that means that traffic is counted in Bytes. However the unit for these items is "bps", which means "bit per second". If I am correct, units should be changed to "Bps" - "Bytes per second". Also see http://en.wikipedia.org/wiki/Kbps
The same applies to other OS templates.
no, it's the other way around 
we get bytes from the device. one byte is 8 bits. we multiply received value in bytes by 8, getting the value in bits
Oh, stupid me!
i guess we can close this one, other problems should go in new issues
Sorry to reply to an older, closed issue
Jens Berthold, where did you find the number mapping between the performance counter names and the corresponding number?
edit: nevermind. Answer here: http://stackoverflow.com/questions/5581859/retrieve-performance-counter-value-in-a-language-independent-way
the mapping is kept here: HKEY_LOCAL_MACHINE\SOFTWARE\Microsoft\Windows NT\CurrentVersion\Perflib\009
zabbix manual lists one method :
https://www.zabbix.com/documentation/2.2/manual/config/items/perfcounters#numeric_representation