Re: GSE IDs [Re: IETF multihoming powder: just add IPv6 and stir]

[Brian E Carpenter <brian@hursley.ibm.com>]

If the prefix is mutable en route, it doesn't matter that it's
globally unique: the end points can't use it as part of any checksum,
so either you have to add HIP, or the 64 bit suffix has to be
unique enough for checksum purposes. (You might be prepared to
accept an RFC 3041 suffix as unique enough, if you think that
network level privacy is more important than the chance of a checksum
coincidence.)

   Brian

David Conrad wrote:
> 
> Noel,
> 
> I agree.
> 
> However, I have been told that ensuring the global uniqueness of the
> last 80 bits is non-trivial because people like the stateless
> auto-configuration stuff that has been defined for v6.
> 
> One way of addressing this particular concern is to say the top 48 bits
> are mutable from one globally unique value (a "site identifier") to
> another globally unique value (an "aggregation locator") via some
> mapping function at both exit from the source site as well as entry at
> the destination site.
> 
> In this case, the identity portion of the name (the first 48 bits
> within a site) is globally unique by itself.
> 
> Rgds,
> -drc
> 
> On Monday, May 5, 2003, at 01:47  PM, J. Noel Chiappa wrote:
> 
> >> From: David Conrad <david.conrad@nominum.com>
> >
> >> All the GSE-like solutions discussed to date that I am aware of have
> >> assumed the first 48 bits (the locator) are globally unique. ... the
> >> last 80 bits may or may not be unique depending on what you want to
> >> do.
> >
> > There's not much point to separating location and identity if the
> > portion of
> > the name which is the identity is not globally unique *by itself*.
> >
> >       Noel
> >