On Sat, Jul 30, 2011 at 4:31 AM, Michel Py
<michel(_at_)arneill-py(_dot_)sacramento(_dot_)ca(_dot_)us> wrote:
Ole,
Ole Troan wrote:
I presume you are arguing that MPLS (6PE) is not native either?
That's a tough one.
What would make me say it is native is: MPLS is a L2/switching animal,
not a L3/routing one. In theory you can bind any L3 protocol such as
IPv4, IPv6, IPX, Appletalk, etc to it. So the MPLS interface is very
similar in some aspects to a real physical interface such as Ethernet or
HSSI. It reminds me of a frame-relay sub-interface in a past life.
What would make me say it is not native is: you can't remove IPv4 out of
the equation. Frame relay does not even know about which L3 protocols it
transports, while MPLS is kinda going the reverse way in the stack: it
uses L3 packets/datagrams to encapsulate and transport "L2 frames".
Here's my take:
- You can have native IPv6 over Ethernet or HDLC or Sonet or any other
L2 technology.
- Saying you have native IPv6 over fiber or copper is incorrect; you
have native IPv6 over GigE over {singlemode|multimode} (*) fiber or you
have IPv6 over Ethernet over GigE over (*) copper (or other examples)
(*) insert the appropriate 802.x standard
- I like the idea of being native requiring the IPv6 to be bound to a L2
interface. The gray area with 6PE is that the interface is logical, not
physical.
- Native IPv6 over a 6to4 or a 6RD or any kind of L3-L3 tunnel is an
oxymoron.
In other words: native IPv6 means:
a) IPv6 has to be bound to a L2 interface.
b) That interface can NOT be a tunnel interface using another L3
protocol such as IPv4.
Up for grabs:
- c1) Is it acceptable to have a structural requirement to use IPv4
(which would mean 6PE is native) or c2) is it a requirement that the
entire infrastructure (in the case of 6RD, the ISP's infrastructure)
supports IPv6 (which would mean that 6PE is not native).
Food for thought:
- If c2) is chosen, I would consider rephrasing a) so it becomes "IPv6
has to be bound to a PHYSICAL L2 interface". Rationale: besides 6PE, are
there any other gray area candidates?
- If one is in the business of writing an draft about "what is native
IPv6", and if one of the draft's goals is to reach -cough- consensus
-cough-, one may consider "forgetting" the 6PE classification
altogether. The one part that is not open for grabs with me is
classifying 6RD as native.
let me try to write all of the above in my own wording to see if I
understand what you mean...
Anything doing translation L3 <> L3 is not native, that's an easy one.
It mean IPv4 running on top of IPv6, IPv6 running on top of IPv4 is
not native. But this easy way of seeing it only affect IPv4 and IPv6,
not all of the others.
As you said somewhere in there, anything attached to a L2 interface is
L3 and in that respect native as long as the transport from that L2
interface do not require other L3 protocols to work?
(of course it gets hairy when you involve MPLS according to Michel's
text above....)
When Jordi says:
"However, for what it matters here, 6rd is native after exiting from the
ISP, same as 6to4 is native after exiting from the 6to4 relay."
That fails the first one, it is _translated_ on L3 and is because of
that not native, even if the transportation of the packet after the
translation is 100% native as in not require another L3 protocol to
work.
I guess it all boils down to, are we talking about end to end native,
or the transportation of the L3 protocol being native?
--
Roger Jorgensen |
rogerj(_at_)gmail(_dot_)com | - IPv6 is The Key!
http://www.jorgensen.no ; | roger(_at_)jorgensen(_dot_)no
_______________________________________________
Ietf mailing list
Ietf(_at_)ietf(_dot_)org
https://www.ietf.org/mailman/listinfo/ietf