All,
This is a good note on which to end the speculative comments. I
would suggest that rather than commenting on the *possible* direction of
Brian Dickson's proposal, that the WG wait for the actual draft
describing the proposal. That way, comments can be constructive and
directed at a concrete specification.
Regards,
Brian
IPv6 WG co-chair
Jim Jackson wrote:
It is interesting that people who are omniscient don't appear to
actually understand what is being proposed. Brian has had to repeat
several times that his proposal does NOT alter the standard
autoconfiguration operation.
While I'd not claim to know whether his proposal has sufficient merit to
be accepted, I have a gut feeling that a universal fixed 64 bit boundary
may cause future problems. I also think choice is good, as long as that
choice does not come at too higher "cost". But the proposal must be
understood before the costs can be weighed.
On Mon, 24 Sep 2007, Mark Smith wrote:
On Sun, 23 Sep 2007 15:42:01 +0100 (BST)
Jim Jackson <[EMAIL PROTECTED]> wrote:
I've been lurking for a while, but this illustrates what I've seen in
quite a few posts on this list and on the ipv6ops list - some people just
know exactly what EVERYONE needs....
I know what my customers want - and I only operate a network because I
have customers who'll pay me to do it. If I don't know what my
customers want, I won't have them because I won't be giving them what
they want. The very large majority of my customers aren't engineers, so
they don't find technical complexity interesting. They just want it to
work, and they want it to work as cheaply as it can.
Do a survey if you like. Ask people (not just engineers)
which option they prefer in any situation, the hard one or the easy
one. The majority will pick the easy one, and it's usually the simpler
one.
My customers might not understand what fixed length
interface ids are. However, what they will understand is that if a 10
minute job could be made 5, and I spend that saved 5 minutes on
something else for them that's also a need, they'll say, "we'll go with
the simpler and cheaper option for this problem."
So what's best for the IETF's end-users? Simplicity, because it's
cheaper to buy, to configure and to fix. Universal fixed length
interface IDs are simpler.
...and so the inference is that everyone MUST have this option.
Yep, in particular when this thread is trying to change 10 years of
specified protocol and implementation. It's too late for that.
IMHO, the only way Brian would be able get this changed would be to use
IPv6 as the basis for IPngng, and start writing IPngng drafts.
--------------------------------------------------------------------
IETF IPv6 working group mailing list
ipv6@ietf.org
Administrative Requests: https://www1.ietf.org/mailman/listinfo/ipv6
--------------------------------------------------------------------
--------------------------------------------------------------------
IETF IPv6 working group mailing list
ipv6@ietf.org
Administrative Requests: https://www1.ietf.org/mailman/listinfo/ipv6
--------------------------------------------------------------------
