>> I think it would be best to go with your alternative, and require
>> vendor OIDs to be registered IANA enterprise numbers, cf.
>> http://www.isi.edu/in-notes/iana/assignments/enterprise-numbers
>> this outsources the registration.
>
>I do not think that we should do >>anything<< to encourage confusion
>between the space of values which are settable through sysctl and the
>space of values settable in an SNMP ASN.1 MIB.

my point was that if we (netbsd) designate "vendor" (aka CTL_VENDOR)
as the top level node in the sysctl mib as where vendors put stuff,
and imply that vendors should use their oid as the second level
(perhaps VENDOR_OID), then vendor foo (oid 9246) will not be able to
conflict with vendor bar (oid 21356) and programs *compiled* to use
one or the other will simply not interoperate (and give wrong
results).

since the sysctls are basically handled via a switch (or switches),
this ought not to waste too much space or time (as it would if the
sysctl mibs were large arrays or linked lists).

local sounds, to me, like local hacks (ala rc.local or /usr/local),
not vendor hacks.

-- 
|-----< "CODE WARRIOR" >-----|
codewarrior@daemon.org             * "ah!  i see you have the internet
twofsonet@graffiti.com (Andrew Brown)                that goes *ping*!"
andrew@crossbar.com       * "information is power -- share the wealth."