Tony Hain wrote:

> Statements in the document like: 'A cellular host SHOULD NOT
> perform Duplicate Address Detection ...' or lack of support for IPsec
> that is justified as too hard for a small device to implement are
> focused on engineering issues for a particular product, not the
> architectural minimum.


I don't think this was exactly as we wanted to see the document.
First, engineering issues are important, but if we would only follow
them, why didn't the draft say MUST NOT implement RFCs of over 5
pages ;-) ? Secondly, there are two main classes of engineering issues:
those internal to the device and not (so) relevant, and those
related to the characteristics of the communications interfaces
which typically do need to be addressed in some manner.

In any case both examples you cite have NOT been done just
in order to save space. Consider the small amount saved by
DAD, or the recommendations to implement something equally
complicated to IPsec in certain cases. Instead, what we tried
to do was to evaluate the specific conditions in cellular
networks and devices, and figure out if the function was
technically necessary in that context. We may have done
our analysis right, or wrong -- and that's the reason we
are having this discussion. As an example, it isn't just
necessary to show that DAD is expensive, you also have
to show that on a particular class of links you don't need
it. I seem to recall some discussion today on DAD not
being necessary in PPP; I don't know what the conclusion
of the discussion will be but that's the kind of result
we are looking for - please work with us on the technical
analysis.

Jari

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