Michel Py wrote:
> Problem is that IPv6 is much more than IPv4 with more bits. Please note
> that this is not a "I told you so" post; I would certainly have opposed
> what I will suggest below.
Agreed.
> As of ID/LOC separation, the sad truth is that we both tried, and we
> both failed. And we're not the only ones or the first ones or the last
> ones to try either.
I specified and implemented ID/LOC separation and it worked.
So, I have some insight on it.
> impossible to deliver. Such as, {cough} PA-based multihoming.
For PA-based multihoming, ID/LOC is not essential. An ID and
multiple locators is just as good as multiple addresses.
W.r.t. IPv6, however, an 8B ID and 8B locators means about half
amount of storage than 16B addresses. That's all.
So, the better solution should have been IPv6 with 8B addresses.
Then, the problem for PA is that, when you say PA, you must
define what P is, as everyone want to be a P.
To me, it is obvious that there should be small number
of Ps, to which all the other entities (including small ISPs)
should be multihomed.
But, it is also obvious that my idea is not politically acceptable
to regional NICs where small ISPs have dominant voting rights.
Masataka Ohta
_______________________________________________
Ietf mailing list
Ietf@xxxxxxxx
https://www.ietf.org/mailman/listinfo/ietf