On Fri., 16 Aug. 2019, 16:59 Fernando Gont, <fgont@xxxxxxxxxxxxxxx> wrote:
On 16/8/19 09:02, Roland Bless wrote:
> Hi Fernando,
>
> see below.
>
> On 15.08.19 at 20:13 Fernando Gont wrote:
>> On 15/8/19 12:27, Roland Bless wrote:
>> [....]
>>> c) given the increasing number of virtual machines and IoT devices 64
>>> bit isn't sufficient, see also the discussion of new MAC address lengths
>> [...]
>>
>> The MAC addresses should have never been embedded in the IID. In fact,
>> that's no longer the recommended way to generate IPv6 IIDs. See RFC8064.
>
> I guess you misinterpreted my statement, since I
> was not referring to modified EUI64s or IIDs at all and I'm fully aware
> of the RFC.
Sorry for that!
> This was just to point to IEEE work on extending the MAC address space,
> showing the need for larger addresses and that 64-bit aren't obviously
> sufficient, see e.g.,
> https://www.ietf.org/proceedings/96/slides/slides-96-edu-ieee802work-0.pdf
My friend John would say that MAC addresses are not addresses: They
essentially are numeric IDs that are not topologically dependent (hence,
not addresses), and that are required to be globally unique (most of the
time, since you also have the U/L bit). The later can be convenient,
also it is not really a requirement per-se.
Assuming you could select a MAC address (well, "MAC ID") and detect
collisions, you probably wouldn't need much more than, say 16-bit MAC
addresses (unless you are really considering putting 65K devices on the
same broadcast segment).
The story behind 48 bits when 10 bits would have done the job because of a maximum of 1024 nodes, plus lots of other good discussion about flat and hierarchical addressing in general.
"48-bit Absolute Internet and Ethernet Host Numbers", by Dalal and Printis, Xerox, July 1981.
From Yogan Dalal's Ethernet History website:
>From that perspective, you might argue that we could have done with
96-bit addresses (not to say 80-bit addresses, or less) -- alignment
being a different business, of course. :-) Or one could do
variable-length addresses
The specific 128-bit value seems to be a compromise between two
proposals
(https://www.lacnic.net/innovaportal/file/2578/1/ipv6-latnic2018-hinden-01.pdf)...
but I won't speak for Bob. ;-)
P.S.: Of course the ship has already sailed. So other than for the sake
of Internet history or the sake of discussion :-), there's no much of a
point in debating about the best address size, because we are not goin
to change it.
--
Fernando Gont
SI6 Networks
e-mail: fgont@xxxxxxxxxxxxxxx
PGP Fingerprint: 6666 31C6 D484 63B2 8FB1 E3C4 AE25 0D55 1D4E 7492
--------------------------------------------------------------------
IETF IPv6 working group mailing list
ipv6@xxxxxxxx
Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6
--------------------------------------------------------------------
