Re: [v6ops] draft-templin-v6ops-pdhost a working group draft?

["Templin, Fred L" <[e-mail]>]

I agree with Fred. The delegated prefix shows up in the routing system the
same as for any prefix, and the host needs to be the steward of the prefix
in the same sense as an ordinary router would. From the outside world,
the behavior of a host with a delegated prefix should be indiscernible from
that of an ordinary router, i.e., even though it acts as a host internally.

This brings up an interesting question, however - should the host respond
to packets addressed to the subnet router anycast address?

Thanks - Fred

> -----Original Message-----
> From: v6ops [mailto:[e-mail]] On Behalf Of Fred Baker
> Sent: Wednesday, November 29, 2017 3:32 PM
> To: Paul Marks <[e-mail]>
> Cc: james woodyatt <[e-mail]>; [e-mail]
> Subject: Re: [v6ops] draft-templin-v6ops-pdhost a working group draft?
> 
> 
> 
> > On Nov 29, 2017, at 3:01 PM, Paul Marks <[e-mail]> wrote:
> >
> > RFC4291 says "a subnet prefix is associated with one link", but a pdhost prefix is assigned to a host, which doesn't really behave like a
> link (there's no neighbor discovery, for example.)
> >
> > So, is a pdhost prefix actually a "subnet prefix" (with reserved addresses), or something distinct?
> 
> I would argue that it is a prefix, in the sense that a prefix advertised in BGP is a prefix.
> 
> > I would advocate for permitting the host to do whatever it wants with the (128-n) bits.  That could include "treat it like a regular
> subnet prefix" or "build a service where every hash(content) gets an IP address."
> 
> I'm not going to argue against that. That said, "every hash gets an address" works for a 64 bit IID if the hash can be expressed in 64 bits.
> It might argue in a direction you didn't intend. And you really don't want to make 128-n work badly for SLAAC, if fielded systems use
> SLACC (I'm told a number of them do).
> 
> I tend to think of the IA_PD prefix as a bundle of /64 prefixes.