[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Tunneling and Transition Drafts

I need a consensus of the WG.

In what I am about to say, someone might possibly feel their toes stepped on. Apologies in advance - and get used to it. I have big feet, and I often put them in my mouth.

In taking over the working group, Kurt and I basically were presented with the following scenario:
- this group is for operational questions (not protocols)
- this group seeks to conform to RFC 1958 (we are not here to bless very lame solution that comes along)
- transition mechanisms are generally out of scope.

As I understand it, v6tc (which was intended to take over all the tunneling stuff, which is to say a large proposition of the transition stuff) is dead in the water. The expectation was that all the tunneling/transition stuff would move there, but the BOF was a non-event. There isn't likely to be a v6tc WG. (Dear AD: if you disagree, this would be a good time to say so).

That said, there are roughly 80 gazillion different transition solutions out there, including at least silkroad, teredo, DSTM, secure tunnels, etc. To deploy IPv6 in a network that is in parts IPv4-only, in parts IPv6-only, in parts dual, and in parts what Jim Bound calls "IPv6-dominant" (which is not a bad term, btw), which I think I would say is ipv6-only in its service but might use IPv4 internally, it seems to me that one needs six things said or defined:

 - "please turn IPv6 on by default in your end systems"

 - "please turn IPv6 on in your enterprise network"

 - "please turn IPv6 on in your access/ISP network"

 - how to tunnel IPv6 through an IPv4-only/dominant network between
   IPv6-only/dominant hosts or networks

   key attribute: static or dynamic IPv6/IPv4 tunnels between tunnel
   endpoints, and appropriate routing through the IPv4 domain to
   determine what the appropriate tunnel endpoint router would be.
   Could be done with a DNS name not unlike a reverse lookup for
   the IPv6 address but delivering an IPv4 "A" record. Could also
   be done through simple routing if the tunnels sty up or use a
   concept like RFC 1793.

 - how to tunnel IPv4 through an IPv6-only/dominant network between
   IPv4-only/dominant hosts or networks

   key attribute: static or dynamic IPv4/IPv6 tunnels between tunnel
   endpoints, and appropriate routing through the IPv4 domain to
   determine what the appropriate tunnel endpoint router would be.
   Could be done with a DNS name not unlike a reverse lookup for
   the IPv4 address but delivering an IPv6 "AAAA" record. Could
   also be done through simple routing if the tunnels sty up or use
   a concept like RFC 1793.

- how to translate IPv6->IPv4 and IPv6-IPv4 between an IPv6-only/dominant
and an IPv4-only/dominant domain

   key attribute: translation system captures DNS lookups and
   re-advertises its own address in some way, so that IPv4 hosts
   seeking IPv6 peers connect to it and it translates, and IPv6
   hosts connect to it and it translates.

Note that I didn't say what kind of tunnel. GRE, MPLS, IP/IP, IPSEC, L2TP, who knows what else - they all must be addressed.

I have a fairly large number of ways to do those - silkroad, teredo, dstm, and a list of others. Each one, when it looks at the others, says "but mine is better" or "my customers want to deploy mine now".

I have no idea how this transition is going to happen if we can't sound a clear signal on how to do it.

Which brings me to my question. I am being asked by various proponents of various things what the working group wants to do with their approach. I need to know what the game plan for this working group is: let a thousand flowers bloom (they can all ask for informational status from the RFC Editor and the probability that the transition will happen in an orderly fashion rapidly approaches zero), or I need a consensus statement from the working group that every single approach on the table will be set aside and the working group will actively work on providing a single solution that meets all those needs that the IETF can look at and say "yes, that will accomplish an orderly transition in a timely fashion and *I*will*support*that*in*favor*of*all*others".

Opinions?