Tony, 

  > I appreciate where you are trying to go, but I think you 
  > are still being
  > blinded to issues because the focus is to centric on 
  > *host*. Yes people
  > are motivated to worry about that problem first, but making design
  > assumptions based on the mobile device on the end of an 
  > air-link being
  > limited to a host will cause problems. The current logic 
  > appears to be
  > 'we don't want to do DAD, so define the link in terms of a 
  > host so it is
  > unnecessary, then worry about a router later'. The host perspective
  > should always be a subset of the demands of a link technology, so
  > starting there will only assure that the link can never be 
  > used outside
  > that scope. A group that is so focused on overhead of bits on the
  > air-link should really be concerned about the impact of 
  > forcing a router
  > to use IPv6 over IPv6 tunneling to get a reasonable service over it.

=> I think there is a bigger misunderstanding here 
that perhaps isn't obvious to one of us (you or me).
What implications do you see (regarding DAD) which
would cause us to distinguish between a host or a router?

In the example I gave to Keith, if you have a router
there are 2 separate links. The same rules about DAD
on the cellular interface would still apply. 
What am I missing?

  > As a specific comment to your note, rather than assume that
  > implementations will end up with an MTU of 1280, simply 
  > state that the
  > interface MTU == 1280. In any case the infrastructure 
  > components will
  > have to respond appropriately to nodes that implement PMTU, so don't
  > assume that it won't happen just because it is not required 
  > for devices
  > that are directly attached.

=> Sure, PMTU was used as an analogy. The difference
of course is that DAD is link specific, so once
you define the behaviour on a certain link, you
know thatit will be interoperable for all devices
with interfaces on that link. 

Hesham

  > 
  > Tony
  > 
  > 
  > 
  > > -----Original Message-----
  > > From: [EMAIL PROTECTED]
  > > [mailto:[EMAIL PROTECTED]]On Behalf Of Hesham
  > > Soliman (ERA)
  > > Sent: Thursday, March 07, 2002 9:58 AM
  > > To: [EMAIL PROTECTED]
  > > Subject: Cellular host draft - where we are and 
  > suggestions for moving
  > > forward.
  > >
  > >
  > > Hi all,
  > >
  > > I think we're having a pretty useful discussion
  > > about this draft which has highlighted a number
  > > of issues that need to be addressed. This
  > > is an attempt to summarise the issues that were
  > > discussed and see if we can find a way forward.
  > >
  > > Please note, this is not to suggest 1, 2 or 3
  > > drafts ...etc, this is an attempt to resolve the
  > > technical issues first, editing them in documents
  > > is another issue that can be handled after we
  > > agree on what we're going to put in such documents.
  > >
  > > Before I list the issues and responses, I'd like
  > > to highlight an important point: This draft attempts
  > > to do 2 things:
  > >
  > > - Define the behaviour of a cellular host residing
  > > on a cellular link (in this case we only know of
  > > 3GPP networks because no one else has defined
  > > the use of IPv6 in their specific link)
  > >
  > > - List the host implementation requirements in
  > > a cellular network (not only 3GPP) based on
  > > the common characteristics of these networks
  > > (p2p channels, BW, resilience over the air
  > > interface ...etc)
  > >
  > > Perhaps the two points above can be separated
  > > in 2 documents, but I'm really interested in
  > > getting an agreement on the technical issues
  > > _first_. Editorial issues are secondary here.
  > >
  > > So the main issues that were brought up so far are
  > > (please let me know if I missed a major issue):
  > >
  > > Issue: What is a cellular host ?
  > >
  > > Suggestion: A cellular host is _any_ host that
  > > has a cellular interface. This would include
  > > low end devices like basic phones and high
  > > end devices like laptops. Low end devices will
  > > generally follow the behavioural aspects of
  > > the draft as well as the minimum implementation
  > > requirements, due to the lack of computing capacity.
  > > However, high end devices will implement whatever
  > > they want, but MUST follow the required behaviour
  > > _on_the_cellular_ interface.
  > > Please note that when it comes to implementation
  > > requirements, the draft never suggests that
  > > a standard MUST NOT be implemented, so everyone
  > > is free to implement more. The draft aims at
  > > the minimum, interoperable implementation.
  > >
  > > Issue: The use of the word 'terminal' is
  > > confusing.
  > >
  > > Suggestion: replace 'terminal' with Host.
  > > This was a mistake anyway.
  > >
  > > Issue: Should DAD be disallowed
  > >
  > > Answer: DAD for the general cellualar
  > > case is allowed in the draft.
  > > In the 3GPP-sepcific case, DAD is not needed because:
  > >
  > > - The link is p2p
  > >
  > > - The link-local address is given to the
  > >   cellular host (it can't reject it).
  > >
  > > - A /64 is given to the cellular host _only_
  > >
  > > - The GGSN will ensure the uniqueness of llA
  > > and will not use the host's /64 to configure
  > > any of its global addresses. The same goes
  > > for site-local if used.
  > >
  > > So DAD is not needed. This is analogous to
  > > a design decision to always assume an MTU
  > > of 1280, PMTU implementation becomes superfluous
  > > if the host will _never_ send a packet larger
  > > than 1280.
  > > I think we should close this issue unless someone
  > > points out a flaw above.
  > >
  > > Issue: Should IPsec be mandated?
  > >
  > > Suggestion: We should mandate AH and ESP for
  > > all hosts, following RFC2460.
  > >
  > > Issue: Should ND be mandated?
  > >
  > > Answer: ND is mandated in the draft.
  > > Certain options are not mandated on cellualar
  > > links (like LLA suboption) because it's not
  > > relevant. NUD is a MAY because there are
  > > other ways of detecting reachability in
  > > these links. If multiple default routers exist
  > > RSs and RAs will detect the failure of one
  > > of those routers.
  > >
  > > Issue: Stateless address autoconfig (2462)
  > > allowed?
  > >
  > > Answer: Yes it's allowed for the general
  > > cellualr case. However for the 3GPP case
  > > stateless address allocation works differently
  > > from RFC2462 (and does not break 3041), so
  > > come parts of RFC2462 are not relevant
  > > (the DAD part again!)
  > >
  > > Comments so far? Are these suggestions/answers
  > > accepatable?
  > >
  > > It took me 2.5 hours to write this email because
  > > of my wonderful mail server, so I probably
  > > won't be able to give quick responses.
  > >
  > > Hesham
  > > 
  > --------------------------------------------------------------------
  > > IETF IPng Working Group Mailing List
  > > IPng Home Page:                      
  > http://playground.sun.com/ipng
  > > FTP archive:                      
  > ftp://playground.sun.com/pub/ipng
  > > Direct all administrative requests to 
  > [EMAIL PROTECTED]
  > > 
  > --------------------------------------------------------------------
  > 
--------------------------------------------------------------------
IETF IPng Working Group Mailing List
IPng Home Page:                      http://playground.sun.com/ipng
FTP archive:                      ftp://playground.sun.com/pub/ipng
Direct all administrative requests to [EMAIL PROTECTED]
--------------------------------------------------------------------

Reply via email to