>>>>> On Mon, 18 Aug 2003 21:00:36 -0400,
>>>>> "Bound, Jim" <[EMAIL PROTECTED]> said:
> The solution that will work for now is make a statement in the
> IETF and in industry IPv6 implementation documentation
> That link-local addresses SHOULD not be
> included in the DNS.
I agree with this.
Regarding the rest of the statements, my opinion depends on the
details:
> that
> link-local addresses SHOULD not be used as an IPv6 address
> type by applications.
> That link-local adddresses SHOULD be
> restricted to IETF protocols on Hosts to perform Neighbor
> Discovery, Stateless Address Configuration, DHCPv6, or other
> operation protocols to bring a Host up on a network. The bottom
> line is link-local address are not usable for applications.
First of all, we need to define "applications" clearly. I believe all
existing IPv6 routing protocol and DHCPv6 implementations are
user-space "applications." Also, some IPv6 stack implements (all or
some part of) Neighbor Discovery as a user-space "application." So,
it is not really clear if we just say "link-local addresses are not
usable for *applications*."
Additionally, with the fact some fundamental protocols such as
routing/ND/DHCPv6 use link-local addresses, network administrators
need a tool to deal with the addresses. For example, they sometimes
need to ping to a link-local address to see if a remote router showed
in the routing table of another router is alive. As an IPv6 network
operator, I sometimes even need to log in to an on-link node by ssh
using a link-local address, when a routing or ND trouble happens. I
can easily imagine that some people dislike this kind of usage, but
this is a real IPv6 life I'm spending today anyway, and ping and ssh
are of course applications.
I personally think it is reasonable to assume that such administrators
(or more generally "advanced" users) have the ability to disambiguate
the correct link when they use such tools (applications). (Router
administrators have to understand the notion of links after all.)
Also, ping and ssh that I'm using, or more concretely, ping6 and
OpenSSH_3.5p1 on FreeBSD 4.8, do not really care about "links." These
applications just use the getaddrinfo function provided by FreeBSD
4.8, and the library function understands the "%" notation to tell the
correct link to the kernel. (Note: the reasy why these applications
use getaddrinfo() is not because they need to deal with link-local
addresses. They use the function to support IPv6.)
I agree that ordinary users that do not have the knowledge about
"links" should not be bothered with the link-local stuff. I also
agree that it is not meaningful to forth applications something (API,
whatever) just to deal with link-local addresses. (But I'm not sure
if this is something that the IETF should declare with the upper-cased
wording.)
At the same time, however, I believe it is convenient or even
necessary for "advanced" users (such as network operators) to have the
ability to use link-local addresses even with an "ordinary"
applications such as ping or ssh. I also believe it is useful to
define the common syntax and the library support (e.g., the % notation
support in getaddrinfo()) to provide the ability.
I also have some personal opinions on the LLMNR or zeroconf usage wrt
link-local addresses, but I won't go further on this for now. I
believe my main points should be covered by what I've said above, and
I'm afraid discussing the zeroconf usage in this context would make
this thread divergent without contributing to the essential point
here.
JINMEI, Tatuya
Communication Platform Lab.
Corporate R&D Center, Toshiba Corp.
[EMAIL PROTECTED]
--------------------------------------------------------------------
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]
--------------------------------------------------------------------
