Re: 6MAN WG Last Call - draft-ietf-6man-multi-homed-host-03.txt

[Brian E Carpenter <[e-mail]>]

On 06/02/2016 06:06, 神明達哉 wrote:
> At Fri, 5 Feb 2016 11:46:23 +1300,
> Brian E Carpenter <[e-mail]> wrote:
> 
>>>   The seeming open points are:
>>>   + whether to move the host-model discussion to an appendix
>>
>> My personal choice is not to move it, but I would certainly do so
>> if that was the WG's preference.
> 
> Okay.
> 
>>> - Section 3.1
>>>
>>>    As described in [RFC4191] and [RFC4861], a Router Advertisement may
>>>    contain zero or more Prefix information Options (PIOs), zero or more
>>>    Route Information Options (RIOs), or a Default Router List.
>>>
>>>   I'm not sure if I can understand the "or a Default Router List"
>>>   part.  If it means 'Router Advertisement may contain...a Default
>>>   Router List', it doesn't make sense since an RA (or any of its
>>>   options) doesn't contain such a list (Default Router List is
>>>   something that a host maintains based on the content of RA).
>>
>> I think you're correct.
> 
> Then I guess the sentence will have to be revised, but I can't suggest
> anything specific as I don't understand the original intent.  Does
> this "Default Router List" perhaps actually means the source IPv6
> address of the RA (which will be maintained in the Default Router List
> of the receiving host)?

I think we should simply delete ", or a Default Router List". That list
is created and maintained by the host, as you say.

    Brian

> 
>>> - Section 3.1: likewise, the use of 'a default router list' in the
>>>   following sentence is confusing, too:
>>>
>>>    [...]  In a multi-
>>>    homed network implementing source/destination routing, the
>>>    interpretation of a default router list or an RIO has to be modified
>>>    with the words "if the source address is in one of the prefixes I
>>>    advertise in a PIO".
>>
>> This is correct, isn't it? It applies to what the host has stored in
>> its default router list and to any new RIOs that it receives. It could
>> be phrased more carefully, I agree.
> 
> Maybe one source of confusion comes from the fact that this paragraph
> starts talking about the content of an RA.  So one possible way to
> improve the text would be to revise it focusing on hosts' conceptual
> data structures rather than RAs that populate these structures.
> 
> On re-reading the text again, I've also noticed a few more glitches:
> 
>    [...] In their
>    original intent, these indicate general information to a host: [...]
>    "you might create or be configured with an address in this prefix", ...
> 
> Technically, "you might create an address in this prefix" is a matter
> of RFC4862, so I think it should be referenced with RFC4191 and
> RFC4861.
> 
> Also, I'm not sure if "you might be configured with an address in this
> prefix" is part of the "original intent".  While it would be usually
> the case that a global address manually configured or by DHCPv6 is in
> a prefix advertised in a PIO (and normally with L=1), I would rather
> consider it operational reality rather than an "original intent" of
> these RFCs.
> 
> I think I can suggest alternate text to address some or all of these
> points, but to do so I'd like to know the answer to the previous
> point.
> 
>>> - Section 3.4
>>>
>>>    There is potential for adverse interaction with any off-link Redirect
>>>    (Redirect for a GUA destination that is not on-link) message sent by
>>>
>>>   Is there any particular intent about why we specially say "GUA" here
>>>   instead of, e.g., just "a global destination"?  For example, does it
>>>   intend to exclude ULAs?  If so, why?  In case of ULAs, while the
>>>   proposed host behavior may be less critical with ULAs in practice, I
>>>   don't see any technical reason why we can't use it for ULAs.
>>
>> To me it's clear that both "GUA" and "global" include ULAs, because ULAs
>> are defined to have global scope. Do you think there is a formal
>> difference between "GUA" and "global"?  If so, is it documented?
> 
> It's probably not so clearly documented, but, for example, RFC4193 has
> this text:
> 
>    It is expected that they would share the same Subnet IDs with
>    provider-based global unicast addresses, if they were being used
>    concurrently [GLOBAL].
> 
> (Here "they" means local IPv6 addresses) To me this sounds like if you
> explicitly say a "global unicast address" it can implicitly indicates
> the address is not a local addresses.
> 
> On the other hand, I (personally) don't feel such ambiguity about the
> term "a global IPv6 address" since, as you noted above, ULAs have the
> global scope.
> 
> Going back to my original comment, I now see there was no specific
> intent in the use of GUA.  Then I suggest revising this
> 
>    (Redirect for a GUA destination that is not on-link) message sent by
> 
> to
> 
>    (Redirect for a global destination that is not on-link) message sent by
> 
> to avoid possible confusion.  The latter doesn't explicitly say it's a
> unicast destination, but I believe it's obvious from the context (the
> destination of a redirect cannot be a multicast address).
> 
> --
> JINMEI, Tatuya
>