[manet] Re: BABEL question

Whether the anarchy, as you term it, matters depends on two things. First, its probability, in particular compared to the probability of other failure modes. No need to be multiple orders of magnitude better than other failure modes, but you do want to be small compared to them. Second, what are the consequences of failure. If a clash takes down the whole network, that’s not acceptable. If one node is cut out just temporarily, probably not a problem, especially if cutouts of that duration also occur for other reasons.

Of course that makes the assessment not just a matter of probabilities. And scenario dependent.

> On 24 Jul 2025, at 23:29, Donald Eastlake <[email protected]> wrote:
> 
> Hi Christopher,
> 
>> On Thu, Jul 24, 2025 at 6:15 PM Christopher Dearlove
>> <[email protected]> wrote:
>> 
>> That gives you a number in the right ballpark. But not the exact answer if you just divide one by the other as the pairs (or the equality of the pairs) aren’t independent.
> 
> Sorry, you are right, I was oversimplifying. But the probablity of
> conflict is still surprisingly hight.
> 
> Donald
> ===============================
> Donald E. Eastlake 3rd   +1-508-333-2270 (cell)
> 2386 Panoramic Circle, Apopka, FL 32703 USA
> [email protected]
> 
>>>> On 24 Jul 2025, at 23:10, Donald Eastlake <[email protected]> wrote:
>>> 
>>> I believe that, due to the so called "Birthday Paradox", If you select
>>> 2**32 (about 4.2 billion) values from a pool of 2**64 values with
>>> replacement, the probability of a conflict among the selected values
>>> 50%. https://en.wikipedia.org/wiki/Birthday_problem The thing is that
>>> to determine whether there is a conflict, you need to compare all
>>> unordered pairs. With N items, the number of pairs is N*(N-1)/2 so the
>>> number of cases of possible conflict goes up with the square. The
>>> number of unordered pairs of 2**32 things is ~2**63.
>>> 
>>> Thanks,
>>> Donald
>>> ===============================
>>> Donald E. Eastlake 3rd   +1-508-333-2270 (cell)
>>> 2386 Panoramic Circle, Apopka, FL 32703 USA
>>> [email protected]
>>> 
>>>> On Thu, Jul 24, 2025 at 5:51 PM Juliusz Chroboczek <[email protected]> wrote:
>>>> 
>>>>> Imagine a prefix like: "2001:db8:f000::/48" where the lower 16 bits of
>>>>> the /64 do not represent a subnet ID but instead identify a host route
>>>>> delegation service.
>>>> 
>>>> I most respectfully disagree.
>>>> 
>>>> I have no objection to defining a flat prefix of MANET interface
>>>> addresses.  However, the prefix must be flat, as anything else is
>>>> gratuitious complexity that doesn't bring any practical benefits.
>>>> 
>>>> (Henning is right, by the way, a /64 is good enough.  If I recall my
>>>> undergrad combinatorics correctly, a /64 gives a vanishingly small
>>>> probability of collision when the total number of nodes is less than
>>>> 4.2 billion.  For a /48, the value is roughly 10¹².)
>>>> 
>>>> -- Juliusz
>>>> 
>>>> _______________________________________________
>>>> manet mailing list -- [email protected]
>>>> To unsubscribe send an email to [email protected]
>>> 
>>> _______________________________________________
>>> manet mailing list -- [email protected]
>>> To unsubscribe send an email to [email protected]

_______________________________________________
manet mailing list -- [email protected]
To unsubscribe send an email to [email protected]