Re: new version of the draft

[Brian Carpenter CERN-CN <[e-mail]>]

Hi, here are some *personal* comments on the December criteria draft.
I hope to provide some more formal CERN comments asap.

5.1.9 Unique Naming

I agree with the authors that this should not be a selection
criterion. Now clearly, if for any proposal, it can be proved
that the lack of unique (binary) names actually breaks the
network than that proposal is no good. (Maybe we need a criterion
"the protocol must not be broken" :-)

For the record I raised my hand against this criterion at Washington.

5.2.1 Addressing

Why isn't this section headed "multicasting"?

I think the discussion should distinguish more clearly "multicasting
for service location" from "multicast applications". The former
is unnecessary and undesirable (e.g. see the IP over FibreChannel proposal
for ARP without multicast) and the latter do not necessarily
require multicast at the IP level.

So (at the most) this is a SHOULD.

5.2.2 Extensibility

Yes. No. My problem is that I don't understand what extensibility
means, so I don't know if I need it. I raised the possible
criterion of "compatibility with multiprotocol routers" and this
was shot down without trace (but I still want it :-). But isn't
it one form of extensibility?  What are the others?

This criterion needs work before I can judge whether it's a MUST,
a SHOULD, or null.

5.2.7 Performance (place holder)

I would express it as "not significantly slower than IPv4 on
the same generation of hardware" and I think it's a MUST.

Regards,
	Brian Carpenter CERN, [e-mail]
			voice [phone], fax [phone]

| This is a personal opinion, and not an endorsement of            |
| PIP/EIP, SIP/IPAE, Nimrod, or TUBA.                              |