RE: [dhcwg] <draft-ietf-dhc-agent-subnet-selection-03.txt>

Longest prefix matching is only needed if the server allows overlapping subnets (such as 20.0.0.0/8 and 20.1.0/16). If the server doesn't allow this, there is no problem.

- Bernie

-----Original Message-----
From: Van Aken Dirk [mailto:VanAkenD@thmulti.com]
Sent: Thursday, October 10, 2002 10:32 AM
To: 'Bernie Volz (EUD)'; 'Ted Lemon'
Cc: 'dhcwg@ietf.org'; Dedecker Hans; Dekeyser Miek; 'Dirk.Ooms@alcatel.be'
Subject: RE: [dhcwg] <draft-ietf-dhc-agent-subnet-selection-03.txt> - Clas sfull ?

If a relay agent supplies a masked IP address only, the scenario only works fine provided the server does "longest prefix matching" on its available pools.
If the server does not start with the pool having the smallest mask, situations where VLSM (Variable Length Subnet Mask) is applied will not converge no ?

-----Original Message-----
From: Bernie Volz (EUD) [mailto:Bernie.Volz@am1.ericsson.se]
Sent: Thursday 10 October 2002 2:59
To: Van Aken Dirk; 'Ted Lemon'
Cc: 'dhcwg@ietf.org'; Dedecker Hans; Dekeyser Miek; 'Dirk.Ooms@alcatel.be'
Subject: RE: [dhcwg] <draft-ietf-dhc-agent-subnet-selection-03.txt> - Clas sfull ?

There is no need to do this. What's the benefit? 

You'd have to configure the relay agent with more information than it currently 
needs. (While in many cases the relay is probably a router and has this 
information, why is it needed?) The server can figure out the subnet without 
the prefix because the server needs both the address and subnet mask length for 
all of the subnets (at least those it will be allocating addresses on). 

Also, what happens if the server and relay agent disagree on the subnet mask 
length? Suppose you wrongly configured the relay to send 20.0.0.0/23 and the 
server is configured with 20.0.0.0/24? 

- Bernie 

-----Original Message----- 
From: Van Aken Dirk [ mailto:VanAkenD@thmulti.com <mailto:VanAkenD@thmulti.com> ] 
Sent: Thursday, October 10, 2002 3:16 AM 
To: 'Ted Lemon' 
Cc: 'dhcwg@ietf.org'; Dedecker Hans; Dekeyser Miek; 
'Dirk.Ooms@alcatel.be' 
Subject: RE: [dhcwg] <draft-ietf-dhc-agent-subnet-selection-03.txt> - 
Clas sfull ? 

Hi Ted, 

As you as you mentioned in the reply, it is indeed exactly the same problem 
as with giaddr. 

But why can this problem not be solved ? After all, the overloading of 
giaddr (communication with the relay agent and address pool selection) was 
defined 17 years ago (RFC951-1985). All networks were classfull at that time 
but today almost all IP systems can operate in a classless environment.. 

As it is a totally new DHCP Relay Information Sub-Option, one can easily 
define two fields: IP address and IP mask or IP address and Prefix length. 

One could eventually reuse the syntax of <draft-ietf-dhc-csr-07.txt> - The 
Classless Static Route Option for DHCPv4. 

Best regards - Dirk 

-----Original Message----- 
From: Ted Lemon [ mailto:Ted.Lemon@nominum.com <mailto:Ted.Lemon@nominum.com> ] 
Sent: Tuesday 8 October 2002 11:32 
To: Van Aken Dirk 
Cc: 'dhcwg@ietf.org'; Dedecker Hans; Dekeyser Miek 
Subject: Re: [dhcwg] <draft-ietf-dhc-agent-subnet-selection-03.txt> - 
Classfull ? 

> I wonder though how the DHCP server can derive a subnet from a single 
> quantity being an IP address. e.g. 20.0.0.0 can be an alias for 
> 20.0.0.0/8 
> or 20.0.0.0/29. 
> So my guess is that the only way the DHCP server can derive the subnet 
> form 
> is based on the leading bits of the address being either class A, B or 
> C. 

The DHCP server is configured with a list of subnets.   It has to be, 
to make address assignment decisions. 

This is exactly the same problem that you have if you use giaddr for 
subnet selection. 

_______________________________________________ 
dhcwg mailing list 
dhcwg@ietf.org 
https://www1.ietf.org/mailman/listinfo/dhcwg <https://www1.ietf.org/mailman/listinfo/dhcwg>  
_______________________________________________ 
dhcwg mailing list 
dhcwg@ietf.org 
https://www1.ietf.org/mailman/listinfo/dhcwg <https://www1.ietf.org/mailman/listinfo/dhcwg>  