Re: [EXTERNAL] Re: [v6ops] Scope of Unique Local IPv6 Unicast Addresses (Fwd: New Version Notification for draft-gont-6man-ipv6-ula-scope-00.txt)

[Fernando Gont <[e-mail]>]

Hello, Bert,

On 6/1/21 01:45, Manfredi (US), Albert E wrote:
> -----Original Message----- From: ipv6 <[e-mail]> On
> Behalf Of Fernando Gont
> 
>> The math in RFC4193 for "uniqueness" considers *only a reduced
>> number of
> uLA-based networks being inter-connected*. So, when computing global 
> uniqueness, you should consider *all ULA prefixes in use*, not just 
> those of networks you are interconnecting. And when you do that, you
> get a very high probability of collisions (~1).
> 
> This is getting unnecessarily complicated, IMO. ULAs are more than
> just link-local, 

Indeed, my draft argues that the scope of ULAs is larger than that of 
link-locals. They key point is that it also argues that scope(ULAs) < 
scope(GUAs).



> because an administrative domain, such as even an
> enterprise net, can use them, throughout that network. Within such a
> domain, the top /48 can be guaranteed to be unique,

How could you possibly guaranteed that if you're selecting the ULA 
prefix with a PRNG? Probabilities don't buy you any guarantees, except 
if P=1 or P=0, which is clearly not the case here.

Additionally, there's a difference between "the ULA prefix is 
probabilistically unique if you connect a limited number of networks" -- 
which is the math in Section 3.2.3 of RFC4193 -- versus being globally 
unique, which implies that the ULA prefix is unique among or ULA-based 
networks, whether you interconnected them or not.



> because the same
> admin computes those 40 random Global ID bits. Not so? Such as, use
> the same PRNG, document the seed, then pick your five random bytes to
> use for each site, from the long random sequence.

RFC4193 does not mandate a specific algorithm, but rather specified the 
requirements for the PRNG (good for the RFC4193 authors!). So, 
documenting the seed will serve no purpose, because other 
implementations might use a completely different algorithm for 
generating the ULA prefix.



> Now you can organize that enterprise net into separate /48 networks,
> or for that matter, even link various geo-separated sites, through
> tunnels, where each site gets one or more of those random /48
> prefixes. The important point being, only route inside that
> enterprise, 

This one seems, already, an indication of prefixes being non-global.



> and use a consistent method to compute the Global ID bits.

An this one is obviously non-enforceable, for multiple reasons, 
including the fact that RFC4193 doesn't mandate any specific algorithm.



> I just object to equating this with link-local. It's way more than
> that.

I don't think anybody has equated ULAs with link-locals. The argument 
has been, instead, that: scope(link-local) < scope(ULA) < scope(GUA)

In a way, the expectation of ULAs to be global-scope is probably what 
drove the recent discussion on the 6man list about a ULA registry -- 
which yes, is probably the only way in which ULAs could actually be GUAs 
(except that unless you can enforce the use of such registry, having a 
non-enforceable partial registry won't make the ULA prefixes global scope).

Thanks,
-- 
Fernando Gont
SI6 Networks
e-mail: [e-mail]
PGP Fingerprint: 6666 31C6 D484 63B2 8FB1 E3C4 AE25 0D55 1D4E 7492