Re: <draft-ietf-6man-rfc4291bis-09.txt>

[Philip Homburg <[e-mail]>]

I find the presentation in Sections 2.4, 2.4.1, and 2.5 very confusing.

Technically, the text is not wrong, but I really wonder if for somebody
new to IPv6 we should explain the address architecture this way.

Section 2.4 starts by explaining that all addresses have a subnet prefix /
IID split. This reinforces the IPv4 notion that address assignment and
router are connected. For extra confusion it describes how some hosts may not
be aware of the split.

Then Section 2.4.1 says "It is recommended that the same interface identifier
not be assigned to different nodes on a link."

Later Section 2.5 says that some addresses that look like unicast addresses
are actually anycast addresses and then Section 2.5 has a lot of confusing
text on how anycast is supposed to work.

In practice, ND just has an 'O' flag which essentially specifies an anycast
address when clear.

Practical suggestion, remove all of the routing talk from Section 2.5,
redefine anycast to be within a single subnet and move 2.5 to be just before
2.4.1

But the main issue is that Section 2.4.1 implicitly talks about SLAAC without
actually saying so. 

And then in the middle of a sentence, it say, oh yeah, if you are doing
manual configuration, all this doesn't apply.

I assume that DHCPv6 IA_NA is considered manual configuration in this 
context. Otherwise there is no sensible way of reading that RFC.

Practical suggestion, in Section 2.4, describe that address assignment is
separate from routing. That onlink prefixes can learned from RA or be
configured manually.

Then that some address assignment mechanisms, such as SLAAC have an
prefix / IID split. Other assignment mechanisms such as manual config
(and RFC 6164) or DHCPv6 IA_NA treat the address as an opague 128-bit value.

(then have the section on anycast)

And then the section that describes how IIDs are used in SLAAC.