Margret has raised several valuable questions about
draft-ietf-ipv6-cellular-host-00.txt, which show it is clearly is not
ready for last call. While I am willing to believe that there are valid
reasons to make adjustments based on characteristics of the link, this
document is fundamentally flawed because it is based around the pretense
that small footprint devices are somehow special. We MUST NOT allow
ourselves to modify well thought out architectural fundamentals based on
point-in-time engineering constraints that are dubious at best. If a
device wants to participate as an Internet node, there are basic
requirements. Of course, participating in the Internet is optional, so
if a device would prefer to avoid the requirements, it may choose to
create its own universe.

The argument that small devices have limited processors or memory
overlooks the fact that the processor and memory of many hand-held
devices today is significantly greater than the workstations that were
available when IPv4 was defined. Interesting point; there was no problem
getting the stack and an array of applications to fit then. Another
interesting point; there are frequently rumors that laptops will include
cellular interfaces, so where does the processing and memory constraint
fit in that case?

On the subject of applications, the absolute BS about limiting the
application set to avoid an IPsec requirement is something that belongs
in a product development discussion, NOT a standards discussion. The one
point that should be clear is that over time the number of applications
used via wireless (cellular or otherwise) will grow, and that we can't
predict what they are, much less what they will need from the stack. To
that point, we MUST reiterate that ***ALL IPv6 IMPLEMENTATIONS MUST
INCLUDE SUPPORT FOR IPSEC***. If we relax that requirement, applications
will never be able to expect support, therefore will have to keep
inventing their own mechanisms. The only way to prevent new applications
from appearing on computing devices is to put the executable code in
non-rewritable, non-replaceable, rom.

If a document on cellular requirements makes any sense, it has to be
based on differences in fundamental characteristics of the air link, not
the device on either end. Avoiding DAD simply due to expense is not a
valid justification, but if the loss characteristics of the link make it
more problematic than valuable, there may be an argument. Keep in mind
that doing so precludes the use of RFC3041 addresses, so the
applications where anonymity is most valuable (as one moves around and
doesn't movement traced) are precluded. A non-zero probability of
collision requires a mechanism to resolve duplicates, and DAD is the
currently defined one. If another one exists that makes more sense over
a lossy air-link, we should consider replacing DAD because it will
probably be more reliable in the general case as well.

Tony


--------------------------------------------------------------------
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