[e2e] Addressing format...network and mask

[Joe Touch]

Alok Dube wrote:

> why do I have to stick to "bit boundaries"?
> 
> they are meant for routers to lookup faster

They are the definition of Internet routes as well.

> ...so as long as i send them the
> "range" let the routing node make its lookup table as it sees fit for
> aligning to its bit boundaries...
> 
> why is that a problem?

Bit boundaries aren't decided by the router; they are a property of the 
route.

The definition of Internet forwarding is based on "longest prefix" - 
which is undefined when there are multiple ranges overlap. Multiple 
bit-boundary masks can't overlap and not have different lengths; length 
determines which route is chosen.

The result (and the problem) is that you end up with nondeterministic 
routing, unless you define a different (non-Internet) tie-breaker, e.g., 
smallest range AND lowest start-of-range. I.e., redefine Internet 
forwarding.

Joe

> 
> -thanks
> Alok
> ----- Original Message -----
> From: "Michael B Greenwald" <mbgreen at dsl.cis.upenn.edu>
> To: "Alok Dube" <alok.dube at apara.com>
> Cc: <end2end-interest at postel.org>
> Sent: Monday, December 01, 2003 10:13 PM
> Subject: Re: [e2e] Addressing format...network and mask
> 
> 
> 
>>   Mon, 1 Dec 2003 13:51:47 +0530
>>   "Alok Dube" <alok.dube at apara.com>
>>
>>   I need a way to propogate the 1st address and last address of a
> 
> network.
> 
>>   Incase i have 129 hosts in a network, i simply should propogate the 1st
>>   address and the last address (or the number of hosts) instead of the
> 
> mask.
> 
>>I'm not sure what you really need to do here, but if I understand
>>somewhat correctly, then why not just allocate two subnets, one with
>>128 hosts and one with 1 host?  If you are trying for a dense
>>allocation of addresses within the net, then, in the absolute worst
>>case, there will be 25% waste.  Usually, it will be much more densely
>>packed (no waste at all in your example of 129).
>>