[manet] Re: BABEL question

Juliusz,

> -----Original Message-----
> From: Juliusz Chroboczek <[email protected]>
> Sent: Friday, July 25, 2025 12:24 PM
> To: Templin (US), Fred L <[email protected]>
> Cc: Henning Rogge <[email protected]>; [email protected] List <[email protected]>
> Subject: Re: [manet] Re: BABEL question
> 
> > The overall sense I get from your previous two messages is that you seem
> > to be focused on individual small-sized MANETs taken in isolation and
> > that never merge with other MANETs.
> 
> I don't see what could have given that impression.

OK, if our big-picture world views coincide that is good - we are therefore both
considering a population size at least as large as the number of people on the
planet. Google "how many cellphones are there in the world" says:

"There are currently more mobile phones than people in the world. While the
exact number fluctuates, estimates suggest there are over 12 billion mobile
connections worldwide"

> > So, the population size should be on the order of 10^10 with aspirations
> > of becoming larger still.
> 
> In my message of 24 July, I said:
> 
> > 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.

RFC4429 and RFC9374 both adequately characterize the collision probability
for a given number of bits of randomness for a given population size. For a
/64 and with 10^10 population size, there is a 1% chance that two nodes
may configure the same randomly-chosen value. In other words, a non-zero
likelihood that a node may somewhere in the world have an identical twin;
and if the two ever come into a common operating environment then there
is a collision with serious consequences.

If you want to talk about /48 for randomness in the case of ULA, then that
effectively extends the "host" part of the address into the "routing prefix"
part. But, ULAs can appear as prefixes with length /64 or shorter in RA
PIOs/RIOs and the host route configured from a /48 could appear to be
on-link with a router that originated the PIO/RIO. A better alternative is
to use an IPv6 prefix specifically reserved for host routes only and not a
ULA also designed for configuring true subnet prefixes.

Thank you - Fred

> -- Juliusz

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