RE: Node requirements: draft-ietf-6man-node-req-bis-03.txt

["Manfredi, Albert E" <albert.e.manfredi at boeing.com>]

> -----Original Message-----
> From: Thomas Narten [mailto:narten at us.ibm.com] 

> How's this to drive the point home further... 2461 (Neighbor
> Discovery) is NOT mandated. It is only listed as a SHOULD. (This is
> because some link layers might not need all parts of ND. But this has
> turned out to be bizarre because the important LLs do require ND, so
> it effectively is a requirement.)

So perhaps some guidance on the attributes of the LL which make ND a
requirement should be included?

> Of course, 2462 is widely implemented, and I'd strongly recommend that
> devices should implement it (though I can think of some
> exceptions). But we do have a fundmanental question about whether the
> IPv6 Node Requirements document can make new requirements that are not
> already in place in existing standards track documents.

It is not entirely illogical to have multiple RFCs that describe
different aspects of IPv6, different possible tools in the toolchest
(such as DHCP, ND, and IPsec), and then one RFC that specifies the
minimal set of these tools that each category of IPv6 node must
implement. Something analogous to RFC 1122, for IPv6.

The title of the RFC, IPv6 Node Requirements, certainly implies that
this is its function. And furthermore, that's how some organizations,
such as DoD, have used the RFC.

To turn the question around, what is the purpose of RFC 4294, if not "to
define the common functionality required from both IPv6 hosts and
routers"? Seems to me that the mandatory and optional ingredients in
IPv6 nodes are likely to change in time, and that it is far easier for a
system designer to look through the requirments in one RFC than it is to
try to interpret the intent of multiple RFCs.

Bert