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

["Woundy, Richard" <[e-mail]>]

I *almost* agree with Bernie's email below.
 
I agree that the Lease Query spec should not force a DHCP server operator to
put reservations into the server, and not force the DHCP server to support
such reservations, solely for making Lease Query work with such
reservations.
 
Although I agree in flexibility in reservation handling, I think we also
need to ensure full multi-vendor interoperability for Lease Query clients
(e.g. relay agents) and servers. That's the point of IETF standardization,
right?
 
A relay agent might choose (for simplicity of implementation) to use DHCP
Lease Query as its sole external source of MAC/IP address information.
Alternatively, a relay agent might choose to use DHCP Lease Query for DHCP
lease information only, and use other mechanisms (e.g. local configuration,
local routing state, remote MIB) for other MAC/IP address information. A
specific, realistic example of another mechanism would be Reverse Path
Forwarding. An access concentrator may be able to verify the source IP
address of an upstream datagram by noting that the source IP address belongs
to a particular IP subnet which has a particular nexthop IP gateway in the
IP forwarding table, and the nexthop IP gateway maps to particular location
information that can be discovered by DHCP Lease Query and/or by local
configuration.
 
We may have an interoperability problem if a relay agent expects only DHCP
lease information in a Lease Query response, but gets other information
(e.g. reservation information) in a response. This will lead to problems if
the access concentrator expects to query other sources of information (e.g.
a MIB) upon the failure of a Lease Query message -- and if the other
source(s) would have responded with different IP/MAC address information
than the (reservation in the) Lease Query response.
 
-- Rich

-----Original Message-----
From: Bernie Volz (EUD) [mailto:[e-mail]]
Sent: Tuesday, March 11, 2003 10:01 AM
To: 'Erik Nordmark'
Cc: 'Mark Stapp'; Steve Gonczi; [e-mail]
Subject: RE: [dhcwg] Leasequery: should it be standardized?



Erik: 

I don't disagree with your and Thomas' concerns, and I probably didn't read
earlier dialogues carefully enough. Ralph and I spoke about this issue at
Connectathon and I now understand the specific concern.

I think DHCP should be used as it is today and the lease query may be used
to extract that information. If the DHCP server has information regarding
client reservations (or whatever you want to call fixed address
assignments), there is no reason it can not report on those exactly as it
would for any other lease. Whether the client uses DHCP to obtain its
information or has static configuration is immaterial - the point is the
that client *COULD* use DHCP and would get the information.

In this case, as DHCP provides the client no indication that this is a
reservation, I don't think lease query should either. So, I would be all for
removing that indication from lease query.

I also think that lease query should not recommend steps to add these
reservations to the server if there is no intention of these clients ever
using DHCP. But, I also feel that lease query should not prevent someone
from doing this - it should simply be silent on the issue as this is a usage
issue and not a standards issue. If a vendor wants to recommend doing this
(and I can see reasons why they would), they could.

- Bernie 

-----Original Message----- 
From: Erik Nordmark [ mailto:[e-mail]
<mailto:[e-mail]> ] 
Sent: Tuesday, March 11, 2003 8:52 AM 
To: Bernie Volz (EUD) 
Cc: 'Mark Stapp'; Steve Gonczi; [e-mail] 
Subject: RE: [dhcwg] Leasequery: should it be standardized? 


> I agree ... we don't want to restart this process. There's a lot of good 
> reasons why DHCP should be used - it is after all DHCP information. The 
> existing framework is fine regardless of whether the devices and 
> applications already knew dhcp. 

Bernie, 

As I understand Thomas' concerns part of the issue is that 
the draft talks about leasequery being used when there isn't a DHCP lease. 
Thus it's expanded to be able to ask questions about MAC/IP addresses 
in general. 

I think that distinction is quite important. 
Arguing that DHCP be used to query information about DHCP leases 
is one thing. 
Arguing that DHCP is the right protocol to use to query about MAC/IP address

related information is a rather different thing. 

   Erik