Re: [renum] SLAAC/DHCPv6 addr-conf operational gaps

["Liubing (Leo)" <[e-mail]>]

Hi, Mark

Thanks for your comment. 
I like the concept of separating "_address configuration_ methods" and "address aging methods", if we could initiate the subsequence work, it should be considered.

All the best
Bing

> -----Original Message-----
> From: Mark Smith [mailto:[e-mail]]
> Sent: Wednesday, February 27, 2013 4:35 AM
> To: Liubing (Leo); [e-mail]; [e-mail]
> Cc: [e-mail]
> Subject: Re: SLAAC/DHCPv6 addr-conf operational gaps
> 
> Hi,
> 
> I've had a quick read though, I'll try to do a thorough one in the next week or
> so.
> 
> Thanks for writing this, I've been thinking about doing something similar over
> the last few months since, IIRC, I heard about Windows 7 invalidating IPv6
> addresses learned via DHCPv6 when SLAAC was enabled and DHCPv6 was
> disabled on a link.
> 
> A few thoughts I've had which may be useful -
> 
> Firstly, I came to realise that the mistake being made by Windows 7 was the
> assumption that the aging of the assigned IPv6 addresses was tightly coupled
> to the DHCPv6 session state, meaning that if DHCPv6 went away, so would
> the assigned addresses. This is generally the way DHCPv4 has worked.
> However, in IPv6/DHCPv6, it seems the address lifetime values in the IA_NA
> are not coupled to the T1 and T2 times, so if DHCPv6 goes away (perhaps
> because the M bit was switched off in a latter RA), the configured IPv6
> addresses should be left to age out as per their preferred and valid lifetimes.
> 
> The other thing I noticed was that the M RA flag and the PIO A flags aren't
> mutually exclusive - I couldn't find any text in RFC4861 that says if the M bit
> is switched on, the PIO A flags must be switched off and vice-versa. So that
> suggests that the DHCPv6 and SLAAC can co-exist, and I think that is quite
> reasonable if you're transitioning a link from DHCPv6 to SLAAC address
> configuration or vice-versa.
> 
> Thinking about it more, I came realise that DHCPv6 and SLAAC are _address
> configuration_ methods, but are not address aging methods - IPv6 takes care
> of address aging via it's preferred and valid aging mechanisms, regardless of
> the address configuration method. Static assignment is also just an address
> configuration method, although typically the addresses don't age out, but
> only because they're normally set to infinity. Perhaps in the future there
> might be other address configuration methods.
> 
> I didn't seem to be able to find an RFC that made it clear that address
> configuration and address aging are quite separate, so perhaps this ID could
> be the one.
> 
> Regarding this text:
> 
> 'For the host behavior, there is an explicit rule in the SLAAC specification
> [RFC4862]: "If the Autonomous flag is not set, silently ignore the Prefix
> Information option."'
> 
> I think RFC5942, "IPv6 Subnet Model: The Relationship between Links and
> Subnet Prefixes." updates that advice, as the PIO option is used to indicates
> which prefix/range of addresses are on-link, via the PIO O bit, even if the PIO
> A bit is switched off.
> 
> Best regards,
> Mark.
> 
> >________________________________
> > From: Liubing (Leo) <[e-mail]>
> >To: "[e-mail]" <[e-mail]>; "[e-mail]" <[e-mail]>
> >Cc: "[e-mail]" <[e-mail]>
> >Sent: Tuesday, 26 February 2013 6:14 PM
> >Subject: SLAAC/DHCPv6 addr-conf operational gaps
> >
> >Hi, 6man & v6ops
> >
> >We submitted a new draft to discuss the SLAAC/DHCPv6 interaction gaps.
> >
> >As we know there are several flags in RA messages regarding with the host
> configuration behavior, which are A (Autonomous) flag, M (Managed) flag,
> and O (Otherconfig) flag.
> >For some reason, the host behavior of interpreting the flags is ambiguous in
> the standard (mainly RFC4862). I presented a draft discussing M flag
> behavior in 6man @ietf84, and there were some feedbacks arguing the same
> issue. This draft analyzed all the three flags, and provided test result of
> current implementations, it showed the behavior of different mainstream
> desktop OSes have varied. The ambiguous and variation might cause
> operational problems, such as renumbering (used to discuss in 6renum WG
> and been documented in the WG drafts), cold start problem, and
> management gaps .etc.
> >
> >Your review and comments would be appreciated very much.
> >
> >All the best,
> >Bing
> >
> >> -----Original Message-----
> >> From: [e-mail] [mailto:[e-mail]]
> >> Sent: Monday, February 25, 2013 5:52 PM
> >> To: Liubing (Leo)
> >> Cc: [e-mail]
> >> Subject: New Version Notification for
> >> draft-liu-bonica-dhcpv6-slaac-problem-01.txt
> >>
> >>
> >> A new version of I-D, draft-liu-bonica-dhcpv6-slaac-problem-01.txt
> >> has been successfully submitted by Bing Liu and posted to the
> >> IETF repository.
> >>
> >> Filename:     draft-liu-bonica-dhcpv6-slaac-problem
> >> Revision:     01
> >> Title:         DHCPv6/SLAAC Address Configuration Interaction
> Problem
> >> Statement
> >> Creation date:     2013-02-25
> >> Group:         Individual Submission
> >> Number of pages: 12
> >> URL:
> >>
> http://www.ietf.org/internet-drafts/draft-liu-bonica-dhcpv6-slaac-problem-
> >> 01.txt
> >> Status:
> >> http://datatracker.ietf.org/doc/draft-liu-bonica-dhcpv6-slaac-problem
> >> Htmlized:
> >> http://tools.ietf.org/html/draft-liu-bonica-dhcpv6-slaac-problem-01
> >> Diff:
> >>
> http://www.ietf.org/rfcdiff?url2=draft-liu-bonica-dhcpv6-slaac-problem-01
> >>
> >> Abstract:
> >>    This document analyzes the host behavior of DHCPv6/SLAAC
> interaction
> >>    issue. It reviews the standard definition of the host behaviors and
> >>    provides the test results of current mainstream implementations.
> Some
> >>    potential operational gaps of the interaction are also described.
> >>
> >>
> >>
> >>
> >> The IETF Secretariat
> >
> >--------------------------------------------------------------------
> >IETF IPv6 working group mailing list
> >[e-mail]
> >Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6
> >--------------------------------------------------------------------
> >
> >
> >