RE: [dhcwg] Leasequery: should it be standardized?

[Mark Stapp <[e-mail]>]

I agree with Bernie and Kevin. I particularly liked the 'data-centric' 
words in Kevin's post. DHCP servers tend to know information about DHCP 
clients and IP addresses, they're the sole authority for that information, 
and that information may be useful to other applications or entities on the 
network. Copying the information into some other database may be possible 
in some cases. Or some devices may be able to watch all DHCP traffic and 
build up their own copies of the data that the DHCP servers know. Or SNMP 
may eventually be available both at the servers and at the queriers. But 
lease-query is sort of an application-level INFORM, I think: it's the 
lowest common denominator way to ask about this useful data. The senders of 
the queries aren't necessarily looking for DHCP address leases, but they 
are looking for information that the DHCP servers know, and the senders 
already know how to read and write DHCP packets.

Lease-query isn't policy - it's not 'enforcement', it's not 'admission 
control', and it doesn't require that the servers or the queriers be able 
to do those things. It's a protocol: it defines bits on the wire. There are 
multiple interoperating implementations in multiple OSs and devices. 
There's been enough interest in it to draw competing vendors to invest 
resources in implementing it. There's been enough interest in it that the 
initial implementors were asked in ietf meetings to bring it into the 
working group and advance it there.

There's no question that DHCP servers know some interesting things, and 
that other processes might use what they know to make sophisticated 
decisions or be more robust or provide other enhanced capabilities. I have 
the feeling that someone saw some 'policy' words in the 'problem 
statement'. The solution to that might be to remove those words, and talk 
only about making data that the DHCP server possesses available through a 
DHCP message. That is something worth doing, and something that is 
reasonably done within this working group.

-- Mark

At 03:03 PM 2/26/2003 -0600, Bernie Volz (EUD) wrote:

>Kim/Kevin:
>
>I agree with Kevin's statements. I too think it is important that this is
>standardize by the IETF (which WG is less critical, though DHC WG seems
>like a good home to me because it is after all DHCP).
>
>I agree with Kevin's point about there needing to be some standard
>means to query the DHCP server's database about clients.
>
>For example, this might also be used by routers or other devices on the
>network to handle admission control. The first packet for an IP address
>that is not in the router's OK list (or hasn't been checked in a while)
>could be confirmed (IP + MAC) via leasequery.
>
>There are also other uses of this, for example for network monitoring.
>Perhaps there are other ways this problem could be solved, for example
>by finishing up an SNMP MIB for DHCP that allows this kind of information
>to be obtained. But, that road has been tried and has not been finalized.
>
>And, finally there are probably future uses of this kind of capability
>that we've never would have thought about. So, setting up something that
>is more flexible and powerful than just solving a very small and specific
>problem is, IMHO, the IETF way.
>
>Perhaps a fix is to state that the original problem statement was blah,
>blah, blah, and leasequery meets those needs, plus more.
>
>- Bernie Volz
>
>-----Original Message-----
>From: Kevin A. Noll 
>[<mailto:[e-mail]>mailto:[e-mail]]
>Sent: Wednesday, February 26, 2003 3:17 PM
>To: Kim Kinnear; [e-mail]
>Cc: Ralph Droms; Thomas Narten
>Subject: RE: [dhcwg] Leasequery: should it be standardized?
>
>
>Kim,
>
>At the risk of sounding foolish since I'm not an active
>participant in the WG...
>
>I think there is a need to standardize leasequery.
>
>It seems to me that the problem statement is the real issue.
>
>LeaseQuery should not be positioned in any way as an access
>control mechanism or *necessarily* involved in such, especially
>since there are other uses for this information.
>
>Leasequery is just what the name implies - a *standard* method for
>external entities (whether they be routers or some other system)
>to query the DHCP server for lease information.
>
>As it is today, there is no *standard* backend DHCP database that
>can be queried. Some use text files, some use LDAP, some use proprietary
>databases, etc. Therefore, the ability to query a DHCP server for
>lease information does not exist except through proprietary methods.
>
>This limits interoperability of the various vendors systems. For example,
>there is no way for vendor X to query vendor Y's DHCP server unless
>they cooperate outside the standards bodies.
>
>This functionality is not limited to routers (or other devices)
>using the information for access control. There are cases (not
>necessarily well documented) where systems external to the DHCP
>server need to find out about a lease for network management or
>troubleshooting reasons.
>
>One example is a helpdesk staffer trying to find out whether a
>device is up or down, but only knows the MAC address of the device.
>How does he/she find out the IP?
>
>Okay, DNS could be consulted, but it isn't authoritative for information
>regarding MAC-to-IP mapping, only NAME-to-IP mapping. There have been
>various schemes to store this information in DNS (e.g. the NAME is easily
>derived given the MAC, etc), but there are all kinds of issues that arise
>with this solution (what if we don't *want* that information in DNS, etc.)
>
>The stable storage statements in the current problem statement aren't
>necessarily relevant, either.
>
>My suggestion is that the problem statement be centered around making the
>DHCP server the *authority* for lease information.
>
>An alternative would be to write a new draft (maybe not within this WG) that
>would define a way to make some existing network database (probably DNS)
>authoritative for this information.
>
>I think we could come up with a better problem statement that
>would get your leasequery draft further down the road.
>
>Anyway, that's my 2 cents worth.
>
>--kan--
>--
>Kevin A. Noll, KD4WOZ                           [e-mail]
>CCIE,CCDP
>Perfect Order, Inc.                             717-796-1936
>
>
>-----Original Message-----
>From: [e-mail] 
>[<mailto:[e-mail]>mailto:[e-mail]]On Behalf Of Kim
>Kinnear
>Sent: Wednesday, 26 February, 2003 12:37 PM
>To: [e-mail]
>Cc: Ralph Droms; Thomas Narten; Kim Kinnear
>Subject: [dhcwg] Leasequery: should it be standardized?
>
>
>Folks,
>
>We have come to something of a impasse on the leasequery draft,
>and I need *your* support if you believe we should continue to
>pursue this draft.
>
>===============================================================
>Without considerable support from the DHC WG, we will halt work
>on the leasequery draft and all attempts to bring this work to
>standard status.
>===============================================================
>
>If you believe that there is any value in standardizing the
>leasequery capability, please at least respond to this list ASAP
>with your positive support.
>
>If you have the time and expertise, please read the rest of this
>email and see if you can offer cogent arguments as to why this is
>work that the DHC working group should be pursuing.
>
>If we don't standardize the leasequery capability, each vendor of
>access concentrators and DHCP products that wish to use this
>approach will then need to work together (possibly in some other
>forum) to try to get their products to be compatible.  Of course,
>it may well be that we are the only folks who see this as a
>useful capability, and so that may not be an issue at all.
>
>Thanks -- Kim
>
>-----------------------  Summary -----------------------
>
>In case you haven't been following the email between Thomas
>Narten and myself, he has been questioning the problem statement
>of the leasequery draft.  Ralph proposed a new problem statement,
>but Thomas feels that this whole capability is questionable.
>
>You are invited to respond to Thomas' arguments, which I have
>distilled as follows:
>
>   1.  Doing anything in the DHC WG like supporting "access
>   control in router type devices" is out of scope for the working
>   group, and doesn't fit its current charter.
>
>   2.  Access control in router type devices is not well enough
>   understood to be sure that:
>
>         a) leasequery is the right solution.
>
>         b) any DHC-based approach is the "right" approach to
>         solve this problem.
>
>   3.  Until we are sure of 2(a), then we should not proceed with
>   this work (I believe that this statement is implicit in Thomas'
>   comments.)
>
>-----------------------  Background ---------------------------
>
>Here is Ralph's proposed problem statement:
>
>    Router-type devices which want to enforce some level of access
>    control over which IP addresses are allowed on their links
>    need to maintain information concerning IP<-MAC/client-id
>    mappings.  One way in which these devices can obtain
>    information about IP<-MAC/client-id bindings is through "DHCP
>    gleaning", in which the device extracts useful information
>    from DHCP messages exchanged between hosts and DHCP servers.
>
>    However, these devices don't typically have stable storage
>    sufficient to keep this information over reloads.  There may
>    be additional information that is useful to the device that
>    cannot be obtained through DHCP gleaning.  The leasequery
>    request message described in this document allows a device to
>    obtain information about IP<-MAC/client-id bindings from a
>    DHCP server.  This information may include currently active
>    bindings, bindings involving previously assigned addresses for
>    which the lease on the address has expired and static bindings
>    for devices that are otherwise configured and not using DHCP
>    for address assignment.
>
>Thomas' concerns center on the second paragraph above, and he says:
>
>    Note, that above is pretty vague and doesn't say what
>    information the access device needs.  It's hard to look at the
>    problem statement and say "yes, I understand the boundaries of
>    the problem" and then "and the solution seems like a good
>    match for the problem".
>
>    Popping up a level, how is it even appropriate for the DHC WG
>    to be doing work on "access control in router type devices"?
>    One can argue that work of this broad a scope is well
>    out-of-scope for this WG (e.g., look at the recently approved
>    charter).  I'm far from clear that work of this scope should
>    be done in DHC or that the problem is well enough understood
>    to conclude that DHC lease query is the right solution or that
>    any DHC-based solution is the right one.  What about routers
>    wanting to do access control that don't use DHC, for instance?
>
>    And note, I'm not raising these issue just to be a PITA. These
>    are questions that I expect that the IESG would ask if I
>    brought the document forward.  Thus, I need to have reasonable
>    responses to those questions.  Otherwise, I can predict the
>    likely outcome.
>
>My response to Thomas was:
>
>    This approach to access control was developed by joint work
>    with the folks building our access concentrators and several
>    of us in the DHCP implementation group.  They found that the
>    functionality delivered to actual users was of sufficient
>    value to those users to be worth the cost of engineering this
>    particular solution.  We supported them in moving the
>    implementation forward.
>
>    The solution was not based on the charter of the DHC working
>    group either then or now -- it was based on a rather pragmatic
>    approach to meeting the needs of users, which it has seemed to
>    do.  In my view at least, it fits within spirit of the DHC WG
>    activities, and was a logical extension of the those
>    activities.
>
>    It isn't a comprehensive approach to any sort of security (nor
>    was it designed to be such) -- it is a supporting piece of
>    technology to one limited form of access control.
>
>Thanks for your interest in the leasequery capability.
>
>Kim
>
>_______________________________________________
>dhcwg mailing list
>[e-mail]
><https://www1.ietf.org/mailman/listinfo/dhcwg>https://www1.ietf.org/mailman/listinfo/dhcwg 
>
>
>
>_______________________________________________
>dhcwg mailing list
>[e-mail]
><https://www1.ietf.org/mailman/listinfo/dhcwg>https://www1.ietf.org/mailman/listinfo/dhcwg 
>

_______________________________________________
dhcwg mailing list
[e-mail]
https://www1.ietf.org/mailman/listinfo/dhcwg