Hi Jinmei,


At 02:36 PM 4/2/2003 +0900, JINMEI Tatuya / =?ISO-2022-JP?B?GyRCP0BMQEMjOkgbKEI=?= wrote:
What was the consensus, if any, on alternatives to site-locals when SL
is deprecated?  In particular, which prefixes will we use for
disconnected or intermittently connected "site"s?  I've checked the
chair's presentation slides and the draft minutes, and (roughly)
followed the very recent discussion in the ML, but I cannot be sure
about it.

We did not reach a specific consensus on this point. The WG was attempting to choose an overall direction, after which we acknowledged that there would be further work to do in either case.

If we choose to deprecate site-locals, we will need to settle on
alternative methods to deal with local addressing, access control, etc.

According to the draft minutes, someone mentioned we can use arbitrary
prefixes for those purposes.  That's true, but we cannot assure the
uniqueness of the arbitrary-chosen prefixes, so I don't see any
essential advantage over the existing fec0::/10 (with eliminating the
"full" usage).

I agree. I think that it would be much better to get registries to allocate unique prefixes for use on local networks. People who are ISP connected might also have the option of using part of their ISP-provided /48, if they are not motivated by ISP-independence.

But, even randomly chosen values (the worst possible alternatives in
many ways) are better than site-locals, IMO.

I can live without site-locals, and, furthermore, I personally want
the wg to stop the endless discussion and to concentrate on more
important -from my personal perspective- issues.

Yes! This is the big motivation for deprecating site-locals. It will allow us to _finish_ IPv6 and move on. Someone may want to work elsewhere in the IETF/IRTF on the issue of scoped unicast addressing, and I would support their work. But, this idea isn't baked enough to include in a protocol that is becoming widely deployed as we speak.

Removing half-baked ideas (even promising ones) from the architecture
is a normal process as specification mature.

Right now, the unresolved issues with site-locals are blocking
progress on the scoped addressing architecture, which is vitally
needed to complete the IPv6 specifications.

So, even if the
idea of "arbitrary-chosen prefixes" is just a compromise (without any
real benefit) to convince ourselves, it's okay for me.  Then I'll vote
for deprecating site-local.

Could someone clarify this point, please?

While it would certainly be better to have a means to allocate globally unique private addresses, the operators present at the IPv6 WG meeting in SF were adamant that even using randomly chosen prefixes for local networks would be superior to site-local addresses, given the complexity and issues associated with site-locals.

Besides, we will need another method for private addressing, anyway,
as you cannot nest sites, and many people do nest access control
boundaries.

Margaret



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