[cgl_discussion] IPv6 MIBs/SNMP requirements/support
jagana at us.ibm.com
Thu Feb 20 16:44:20 PST 2003
>Well, the ipv6compliance section in the IPV6-MIB file specified that
>"An agent is not required to provide write access to this object".
Let me clarify this further. For ipv6 "compliance" of mib objects,
it is quite OK to implement at a minimum (i.e. read-only MIN-Access)
the compliance mib object group and however, if you want to
provide ipv6 "conformant" mib object support, then it needs
to support all of them with MAX-ACCESS (i.e. read-write):
At a minimum, we can choose to implement read-only access to begin
with and then extend to provide read-write access to the allowed
objects. And I am quite OK with this.
btw, if you look at more carefully snmp.h file, the ipv6 kernel has
implemented only ipv6IfStatsTable object group only, i.e.
struct ipv6_mib. It needs to be expanded or create new structs further
to support additional object groups(ipv6IfTable, ipv6AddrPrefixTable,
ipv6AddrTable, ipv6RouteTable, ipv6NetToMediaTable under IPv6 General
Since most of these object groups other than few, as you have identified,
are already supported by net-snmp, this support needs to be provided
in the kernel.
Randy, since you are looking at kernel side, you may want to think about
how to accomplish this.
<yixiong.zou at intel.com> To: Venkata Jagana/Beaverton/IBM at IBMUS
Sent by: cc: cgl_discussion at osdl.org
cgl_discussion-admin at lis Subject: RE: [cgl_discussion] IPv6 MIBs/SNMP requirements/support
02/20/2003 03:38 PM
> -----Original Message-----
> From: Venkata Jagana [mailto:jagana at us.ibm.com]
> Sent: Thursday, February 20, 2003 3:18 PM
> To: Zou, Yixiong
> Cc: cgl_discussion at osdl.org; 'Venkata Jagana'
> Subject: RE: [cgl_discussion] IPv6 MIBs/SNMP requirements/support
> >One more thing, do we need to implement those "set" functions?
> From a MIB definition stand point, it is always 'required'
> (unless it says optional) so that it gives the flexibility
> to control these parameters from remote mgmt station. However,
> in most cases, customers won't implement the set capability
> but that doesn't mean it shouldn't be implemented. By default,
> the 'set' functionality can be disabled and however, can only be
> enabled if desired.
> >Most of those defined in IPV6-MIB are not "required" by the MIB.
> Don't think so :) Did you see any stmt from the spec that they are not
Well, the ipv6compliance section in the IPV6-MIB file specified that
"An agent is not required to provide write access to this object".
They also defined that the "Min-Access == read-only".
Yixiong Zou (yixiong.zou at intel.com)
All views expressed in this email are those of the individual sender.
cgl_discussion mailing list
cgl_discussion at lists.osdl.org
More information about the cgl_discussion