Excuse me. This seems brain dead. I think the intent of 1597 was to say here are some numbers you can go play with privately. Since they are private they should never appear on the public Internet, anywhere, ever. If any router on the public Internet sees a packet to (or from) a 1597 network it can throw it in the bit bucket and forget it. It would be possible for some private experiment to use some 1597 addresses to exchange packets over a "wire". Such a wire could be implemented by some pretty complicated arrangements with conventional public Internet service providers -- but any packets with 1597 addresses would have to be encapsulated inside packets with acceptable addresses for the public Internet to go through public Internet exchange points. --jon.
Date: Mon, 30 Jan 1995 16:15:23 -0800 From: postel@ISI.EDU (Jon Postel) This seems brain dead. I think the intent of 1597 was to say here are some numbers you can go play with privately. Since they are private they should never appear on the public Internet, anywhere, ever. Actually, the conclusion that I would draw from this (and that others have certainly drawn) is that RFC 1597 is brain dead. I'd certainly recommend that anyone who was contemplating numbering their network *not* use one of the 1597 numbers, and if their network provider tries to force them to do this "in order to save address space", to go find another network provider. As you've pointed out, by using the 1597 numbers, you're guaranteeing that that you can never use those numbers on the public Internet, and you'll never be able to use more than one network provider. Even if most of a companies hosts are behind a firewall, they'd still be better off using globally unique numbers. After all, if we all do our jobs right, maybe someday we can make the firewalls go away.... - Ted
As you've pointed out, by using the 1597 numbers, you're guaranteeing that that you can never use those numbers on the public Internet,
That's generally the point.
and you'll never be able to use more than one network provider.
Huh?
Even if most of a companies hosts are behind a firewall, they'd still be better off using globally unique numbers.
Yeah, there are 2^32 bits of address space, after all.
After all, if we all do our jobs right, maybe someday we can make the firewalls go away....
If we all do our jobs right, it won't matter if someone uses 1597 space. Firewalls will never go away -- they're too useful. Regards, -drc
This seems brain dead. I think the intent of 1597 was to say here are some numbers you can go play with privately. Since they are private they should never appear on the public Internet, anywhere, ever. If any router on the public Internet sees a packet to (or from) a 1597 network it can throw it in the bit bucket and forget it.
As long as you say "can" rather than "must". Right now these sort of packets will follow the default route. Dino
From: postel@ISI.EDU (Jon Postel) Subject: Re: Test Route Excuse me. This seems brain dead. I think the intent of 1597 was to say here are some numbers you can go play with privately. Since they are private they should never appear on the public Internet, anywhere, ever. If any router on the public Internet sees a packet to (or from) a 1597 network it can throw it in the bit bucket and forget it. What is public and what is private? The only thing that private can mean given the text in 1597 is that it doesn't have global significance. If I make an agreement with a "public" SP to use 1597 addresses, or if two SPs to use 1597 addresses, by your definition, they are not private, but I claim that by local agreement, they certainly ARE private and fulfill the definitions of 1597, if not perhaps the original intent. It would be possible for some private experiment to use some 1597 addresses to exchange packets over a "wire". Such a wire could be implemented by some pretty complicated arrangements with conventional public Internet service providers -- but any packets with 1597 addresses would have to be encapsulated inside packets with acceptable addresses for the public Internet to go through public Internet exchange points. Obviously given the difference of opinions, we're not going to see the following corallary in the same light, but... An IX, and in particular, an RA-administered NAP, should support whatever routing policy has been contracted between the RA and the customer(s) of the RA. If two customers routing policy include sharing a 1597 network, that is something that needs to be considered.
participants (5)
-
David R Conrad
-
Dino Farinacci
-
Paul Traina
-
postel@ISI.EDU
-
Theodore Ts'o