[1781] in North American Network Operators' Group
Re: Policy Statement on Address Space Allocations
daemon@ATHENA.MIT.EDU (George Herbert)
Wed Jan 31 19:21:36 1996
To: Dennis Ferguson <dennis@ipsilon.com>
Cc: George Herbert <gherbert@crl.com>, Sean Doran <smd@cesium.clock.org>,
asp@uunet.uu.net, nanog@merit.edu, gherbert@crl.com
In-Reply-To: Your message of "Wed, 31 Jan 1996 15:28:38 PST."
<199601312328.PAA00272@mailhost.Ipsilon.COM>
Date: Wed, 31 Jan 1996 16:08:20 -0800
From: George Herbert <gherbert@crl.com>
>> Option 5: Provider X can announce nothing outside the area, except
>> to people who are paying X for transit to all X-reachable sites
>> and networks.
>>
>> This would work great if all the backbones touch down in the area.
>> Customers out in the Rest of the World get transit through their
>> backbone to all the area sites. Other regional networks or areas
>> get transit to it via whomever they get global transit from.
>
>Which works fine as long as only one area on the planet ever implements
>Option 5. Any pair of such areas without internal connectivity won't
>be able to talk.
>Seems like a scaling problem.
No, they can talk just fine. In the worst case with asymetric paths,
but they can talk fine.
Backbones A and B, small providers a in region 1 and z in region 2.
a buys transit to the world from A and is in the region 1 block.
z buys transit to the world from B ans is in the region 2 block.
a->z goes a->A. A knows about region 2 block, sends it to region 2.
Once it reaches there, it either goes A->B->z or A->z if there is a
direct interconnect somewhere. z->a goes out the z->B pipe, and
thence to region 1 and either B->A->a or B->a if there is a direct
interconnection somewhere.
-george
