Re: [v6ops] A proposal for draft-ietf-6man-rfc4291bis-07 - timeline

[Alexandre Petrescu <[e-mail]>]

Hi,

Thanks for the pointers to the earlier discussion.

Le 08/03/2017 à 22:54, Timothy Winters a écrit :
> Hi James,
>
> Since this was added in the update from RFC 2461 to 4861 I went to
> go look for why this was added and found the following thread.
>
> Discussion:
> https://mailarchive.ietf.org/arch/search/?email_list=ipv6&q=Requirement+for+64bit+I%2FF+ID&so=date&gbt=1&index=rJtLf5Krh0X9vg3vYts_xO1oUCw
>
>
>
>
Final Decision:
> https://mailarchive.ietf.org/arch/msg/ipv6/IfWdv79U0k9PbwK6-lAIk4yeyK4

I think that decision is that ND should work with arbitrary plens, not
just 64.

In light of that ND decision, it is strange to see an RFC at the same
time IPv6 Addressing Architecture mandating IIDs 64 (and implicit
mathematical plen 128).

Retrospectively:

At time of that discussion, year 2004, there was no IPv6 on cellular
links, [I-D]. Since then 3GPP got many recommendations from IETF about
how to run IPv6. The result is 3GPP User Equipment links only support a
single 64 today.

My supposition, and many indications converge, is that the 64
recommendation from the IPv6 Addressing Architecture RFC led to this
undesirable situation of a single 64 for an UE.

If we were again in 2004, I would tell that decision maker: not only ND
should avoid the 64 limit, but the IPv6 Addressing Archi should too.

That would have saved us many recent RFCs discussing 64 (the BCP CIDR
IPv6, the multiple-address per Host recommendation, the why64 RFC, and
many others I mistakenly forgot).

We are now in 2017...

Alex