RE: [RAM] 6to4

["Templin, Fred L" <Fred.L.Templin at boeing.com>]

Dino,

I'm not sure on all your points, so let's walk through this: 

> > I am wondering where 6to4 fits on the LISP continuum.
> 
> Well, it can work serially. Here is the scenario I am referring to:
> 
> o An IPv6 packet is received by an ITR.

OK; either that, or the ITR resolves a FQDN and gets back a
6to4 address.

> o The DA is an EID so a mapping lookup is done to yield a locator in  
> the 2002::/16
>    range.

OK.

> o The ITR prepends a IPv6 header.

Huh? 6to4 prepends IPv4 headers; not IPv6.

> o The platform specific code in the router will try to 
> resolve a next- 
> hop and find
>    out there is no IPv6 next-hop for the outer DA. So it 
> resolves the  
> address to
>    an IPv4 address which follows the 2002 bits. The ITR prepends an  
> IPv4 header.

So, now you have IPv6-in-IPv6-in-IPv4? 6to4 is only for
IPv6-in-IPv4.

> o The ITR sends the packet out the interface where the IPv4 DA is  
> resolved to.

I don't know what you mean by this; my understanding is
that the ITR would simply launch the IPv6-in-IPv4 packet
into the network and let IPv4 routing steer it to the
correct ETR. 

> > Correct me if I am wrong, but with 6to4 the identifer
> > to locator mapping comes from a single DNS lookup. The
> > identifer is not routeable over the core, but the ETR
> > locator is embedded in the identifier address so the
> > ETR discovery comes "for free".
> 
> The EID in DNS could be a 2002::/16 prefix, but it shouldn't assume  
> the address
> itself or the IPv4 embedded in it is routable.

Not the way 6to4 works; in 6to4, the embedded IPv4 address
is indeed routable.

> > Is 6to4 an example of LISP 2? Is it good enough to be
> 
> Well you can put 2002 addresses in DNS. So from a LISP point 
> of view,  
> it just a
> mapping lookup on a string of bits.
> 
> > considered as a long-term solution, or is it just a
> > baby-step to a fully jacked-up LISP deployment?
> 
> If you used IPv6 addresses as EIDs and IPv4 addresses as locators,  
> then the 6to4 step would be an extra one you would not need. And I  
> wouldn't want to use it because it adds an additional 20 
> bytes to the  
> packet (as described from the sequence above).

AFAICT, there is no difference between LISP and 6to4 in this
regard. LISP encapsulates IPvX-in-IPvY, but for 6to4 it just
so happens that X=6 and Y=4.

Thanks - Fred
fred.l.templin at boeing.com

_______________________________________________
RAM mailing list
RAM at iab.org
https://www1.ietf.org/mailman/listinfo/ram