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

[Philip Homburg <[e-mail]>]

>> Basically a 'host part' doesn't have any properties.
>
But that's not a distinction. I can just as easily give a structure to an I=
>ID as I can to any non-64-bit IID, if I so choose. Or, use a pseudo random =
>value if I'm worried about privacy or security.

I cannot parse this. 

>
>> You can say it identifies an interface, but that is not the case if
>> a 'host part' is part of an anycast address.
>
>Not sure why you make this distinction. Even in the anycast address, these =
>last bits identify an interface, even if one of several possible. It's the =
>topologically closest interface, and others would be identified by the same=
> bit pattern. It's still an ID of an interface, as opposed to being a prefi=
>x.

Maybe I spend to much time with math, but for me an anycast address identifies
a set of interfaces and packet destined to such an address ends up on any of
them. 

To me that is completely different from the case where an address identifies
at most one interface and you know that if it gets delivered to the destination
it will always be the same interface.

>> So let me call them 'host part'
>
>So we would be renaming IID in every RFC. Quite a change. And it's not "hos=
>t" anyway, right? Those bits identify an interface, not a host. So call it =
>"interface part," or "interface ID."

Possibly. Thought it seems that RFC 4861 and 4862 use IID only in the context
of SLAAC so that would not change.

And as a I said before, I consider a set of interfaces something quite
distinct from a single interface.

>I just don't see why IIDs used in SLAAC should be thought of differently. I=
> mean, the anycast example was always a little weird, but those are a tiny =
>minority anyway.

Because way too many seem hopelessly confused. They read about 64 bit IIDs
and do not realize that this only applies to SLAAC.

People write code that if you configure an address manually, the system
automatically considers the /64 prefix onlink.

People managed to write code that drops an onlink prefix because it doesn't 
match the length of a SLAAC IID.