[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
snmpconf to-do lists - deadline April 6
For those of you unable to be with us last week, we can
report that we had two very productive sessions. We now
believe that we are on the downhill slope to completion of
the three documents. Minutes of the meeting will be appearing
Real Soon Now.
At last week's meeting, it was agreed that we would try to
compile the "complete" to-do list for the three documents by
April 6 (this year :-)). In order to do so, we need your help.
Firstly, the current list of unfinished business is below. It
is made up of unfinished items from the list published Jan 12
plus items decided upon during the working group meeting.
Now, you have your chance. Please look through the list
and verify that those things that you want addressed are on
the list. If not, please submit your issue to this mailing
list _no later_ than April 6.
Note that submitting an item does not automatically mean
it'll get done. As the proponent of change, the onus is on
you to convince the working group of the necessity of your
proposed change. Also, you don't have to propose a solution,
merely convince the group that there's a problem to solve.
Jon & David
Current To-do List
Policy MIB Module
#01 Language versioning
Issue: Wes Hardaker raised his concern that we need to have
mechanisms in place for dealing with future revisions to the
language given the fact that it's very unlikely that we'll
never update the language. He will work with Steve
Waldbusser on this.
#02 Bob Moore's questions
Issue: on March 14, Bob Moore published mail with the
subject "Policy MIB: assorted other comments". There are a
number of questions/comments in that mail that need to be
#03 Policy failure and fallback
Issue: if you have multiple policies running at different
levels of precedence, how do you fallback to a policy of
lower precedence if the one at a higher level fails. Steve
Waldbusser had a proposal, which he will work on more with
Wes Hardaker and publish to the mailing list.
#04 Code Reuse
Issue: Wes Hardaker believes that some things should be
done to aid in code re-use. He will publish any proposals
to the mailing list.
#05 Question of notifications and controls for their generation
wrt policy failure.
Issue: should notifications be sent when a policy fails?
If so, what mechanisms should be used to control their
generation? (If I remember correctly,) Wes will work on a
#06 counter32Delta and counterRate review and discussion
Issue: When Steve presented counter32Delta and counterRate,
there was a fair bit of discussion on whether they were
right as they are. More thought and discussion is required.
#07 Dave H. asked for enabled versus admin status for
Issue: David Harrington thinks that the object
pmPolicyAdminStatus might be confusing given its similarity
to ifAdminStatus (but the enums are different). He
suggested changing it to pmPolicyEnabledStatus.
#08 Should we create a create API versus a specific createAndGo
and a specific createAndWait
Issue: at the previous IETF, discussion began having an API
that could use the createAndGo paradigm of RowStatus rather
than createAndWait. Steve Waldbusser presented a proposal
for what this might look like. Subsequent discussion
centered on the idea of having a _single_ API that works for
both createAndGo and createAndWait but hides which it uses
from the caller. Given that we have two and might soon have
three (the EOS WG is also looking at this issue), such an
API would be useful.
Steve will receive assistance from Randy Presuhn and Wes H.
#09 Discussion of permanent storage and configuration update
implications of the Policy Module on the rest of the system
Issue: in DiffServ MIB discussions, the use of
the StorageType TC has been called into question.
Specifically, implementors said that the granularity used
in the DiffServ MIB (storage per row) was not realistic in
an environment where flash was used. We need to look at
how we use StorageType to determine if it's appropriate
for such environments or if we need to come up with an
alternative approach (perhaps a "bigger button" at a
higher level of granularity).
#10 setCli example needs to be written
Issue: given the rules outlined in section 8 (Accessor
Functions) and section 9 (Base Accessor Function Library),
David Partain and Chris Elliot (and any others) will write
an example of how a vendor might add its own setCli accessor
#11 Helper objects for network management system fall over.
Issue: do we need some way of guiding the fallover from one
management system to another? Steve Waldbusser brought this
up long ago and he will either let this drop or will publish
#12 Taxonomy discussion updates to be in BCP referenced in
Issue: the BCP has a discussion of the taxonomy (to
be updated). A non-normative reference to this needs to
be added to the PM module.
Jon will offer a proposal.
#13 Notification Registration Table.
Issue: why isn't the SNMPv3 infrastructure sufficient for
our needs? Steve Waldbusser will publish his reason.
#14 Text on resource usage and clean up.
Issue: although resource usage is largely an implementation
issue, there are some things that we need to do to help.
In particular, Steve Waldbusser noted the problem of
cleaning up fully-created rows when scripts terminate
abnormally. He suggested a registration mechanism for rows.
Steve will present this more fully.
In general, the text needs to be gone over with an eye to
#15 Relationship to other MIB Modules
Issue: some text needs to be written explaining what the
relationship is between this document and other MIBs.
Basically, it would say that the scripts operate on MIB
variables defined in other MIBs.
Jon will offer a proposal.
#16 Table indexing on pmRoleTable
Issues: the IMPLIED keyword serves no useful purpose in
this case and will be removed. Secondly, the index can
become larger than the 128 subidentifier limit set by the
SMI. Consequently, the indexing needs to be re-examined.
#17 The security question(s)
Issue: is the current security model sufficient?
"When executing this action, if SNMP requests are made to
the local system, access to objects is under the security
credentials of the the requester who modified the most
recently modified pmPolicyCodeEntry associated with either
the pmPolicyFilter value or pmPolicyAction value. In other
words, modification of any part of a policy's filter or
action will change the credentials stored for the policy."
Should we consider adding an OwnerString? Some thoughts
from Wes H and Randy Presuhn in this regard would be very
#18 Update capabilities table as per suggestion and add new
read-write override table
Issue: the capabilities table appeared to offer severe
challenges to implementors. So, after much discussion,
the following was proposed:
will all be removed, leaving only pmCapabilitiesIndex and
pmCapabilitiesType. The table will be read-only and will be
populated "automatically" by the agent system. That which
is put into the table will be (1) OIDs from a list of
registered OIDs as suggested by Steve W in mail prior to the
meeting, (2) an OID representing a conformance level with
respect to a MIB module, or (3) the base OID of the MIB
Secondly, a new read-create table will be added called the
pmCapabilitiesOverrideTable. This will be essentially
the same as pmCapabilitiesTable with: Valid or Invalid
state object, and a RowStatus object. This will:
- Allow for manager correction of 'false negative' and
'false positive' reports by the agents registering
- Allows manager to turn off a capability.
- Always survives across reboots and entries in this table
Jon will send a specific proposal for change.
#19 Removal of scratchpad variables
Issue: sometimes you want to clean the slate prior to
using scratchpad variables. Steve Waldbusser presented
a proposal at the meeting, which he will propose to the
#20 Scheduling of policies
Issues: after input from Bob Moore, we no longer believe
that our schedule table is the best way to accomplish what
we want. Steve W. suggested using the scheduling semantics
in the PCIM (policy core info model) done in the policy
framework WG. A specific proposal will be given by Steve.
#21 Examples throughout language discussion
Issue: David Partain believes there are many places where
simple examples would be helpful in understanding language
constructs, MIB objects, etc. He will propose these to the
editors of the document.
#22 Editorial Review
Issue: as we are rapidly approaching finalization of the
document, NOW is a very good time for all to read the
document and get their comments to the list.
#01 The taxonomy discussion needs to be updated if comments are
#02 Correct any remaining technical errors
- On some items, improve guidance
* overlapping controls
* fate sharing on read-create tables
- Agent/X considerations
- Protocol details - Wes & Bert's comments
- Improve Set Latency Example
#03 Correct remaining prose
#04 Finish off MIB Module
Issue: there have been a number of comments on the example
MIB module included in the document. These issues need to
- DisplayString vs SnmpAdminString
- Bob Moore's items
* Agent Caps
DiffServ Policy MIB
#01 Synchronization with the DiffServ MIB
Issue: the DiffServ MIB is going to be re-published
shortly. At that time, the DiffServ Policy MIB will need to
by sync'ed with that document.
#02 Cloning issues
Issue: Once the DiffServ MIB has been published, we
hopefully will know how they intend to deal with cloning of
templates. We may need to add objects to our MIB to do
things the right way.
#03 Example cleanup/additions
Issues: the example may need some work, and there might be
good reason to add another example, including policy script
#04 Name of MIB
Issue: Should this rightfully be called the DiffServ
Policy MIB or DiffServ Configuration MIB or something else?
David Partain will offer a proposal.