Hi, a final reply from me, now that things really get ridiculous...: On Wed, Feb 12, 2003 at 02:47:39PM -0800, Michel Py wrote:
Gert Doering wrote: (And because it makes maintenance and setup much easier,
We call that troubleshooting with the seeds of the pants, which exactly what serious operators don't do.
Serious operators don't reject things that can aid in network setup and management just because some ivory tower research declares it isn't "pure". [..]
We *have* the address space, No we *don't*.
Of course we have. The *space* is there. There are some rules in the way, but that doesn't mean the space suddenly disappears.
The IETF standards are decided by consensus, and the IID bits are not yours to play with for routing purposes. There are other legitimate uses for these 64 bits, such as embedding some crypto in the IID or privacy extensions.
Privacy extensions? On a *point-to-point* link? Now your're getting *really* ridiculous. We're not talking about customer access (they can have a /64 or a /48 just fine) but about backbone lines or tunnels. As for the crypto-in-the-IID - I don't see this happen in the near future for point-to-point links (it might well happen for end-to-end connections), so I'm not very much worried about this.
Of course it is very much possible with 128 bits. People *do* this, so it's possible, isn't it? I will laugh when you have to renumber because suddenly a new security protocol that uses 56 of the 64 IID bits is required.
Required by whom? If *we* decide to implement "some new security protocol" on *point-to-point links* in *our network*, we will have lots of work to do anyway with router deployment tests, router upgrades and so on. This is not going to "suddenly happen upon us". End-to-End security (for systems sitting in some LAN, happily using /64s) is not affected by what the transport layer in between is doing, as long as the transport layer isn't changing packets (which it isn't doing).
The thing that is not possible is to accommodate for that inside the very narrow-minded "one size fits all" mind-set that made the rule that one should use a /64 on a point-to-point link.
This is called "standardization". That's why you can plug a refrigerator, a microwave oven and a cell phone charger in the same power outlet. The microwave takes 10,000 times more power than the cell phone charger, and indeed the wiring for a power outlet that could accommodate only cell phone chargers would be a lot lighter. Do you want 20 types of power plugs in your data center?
Bad example. To use your model: actually we are *not* using 1000 Ampere plugs for everything just because the "same plug for everything". It's ok to standardize certain things - I don't object here - but it's tremendously helpful if standards reflect reality.
There is no reason (except "one size fits all") *for* that rule - or at least nobody in this discussion named one - but many good reasons *against* it.
This has been debated before and the IETF has decided otherwise.
From what I read in Randy's e-mail, I'm not overly convinced that there is/was overly strong consensus here.
Gert Doering -- NetMaster -- Total number of prefixes smaller than registry allocations: 56029 (55671) SpaceNet AG Mail: netmaster@Space.Net Joseph-Dollinger-Bogen 14 Tel : +49-89-32356-0 80807 Muenchen Fax : +49-89-32356-299