Re: [Softwires] [BEHAVE] What is a site? [Re: Some Thought about theAutomatic Tunnel Address

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

Hi Brian,

> -----Original Message-----
> From: Brian E Carpenter [mailto:brian.e.carpenter at gmail.com]
> Sent: Tuesday, September 29, 2009 3:01 PM
> To: Templin, Fred L
> Cc: Dong Zhang; softwires at ietf.org; behave at ietf.org; eric.burgey at orange-ftgroup.com; grobj at ietf.org
> Subject: Re: [BEHAVE] What is a site? [Re: [Softwires]Some Thought about theAutomatic Tunnel Address
> 
> On 2009-09-30 10:10, Templin, Fred L wrote:
> > Brian,
> >
> >> -----Original Message-----
> >> From: behave-bounces at ietf.org [mailto:behave-bounces at ietf.org] On Behalf Of Brian E Carpenter
> >> Sent: Tuesday, September 29, 2009 1:05 PM
> >> To: Dong Zhang
> >> Cc: softwires at ietf.org; Templin, Fred L; behave at ietf.org; eric.burgey at orange-ftgroup.com
> >> Subject: [BEHAVE] What is a site? [Re: [Softwires]Some Thought about theAutomatic Tunnel Address
> >>
> >> On 2009-09-29 19:28, Dong Zhang wrote:
> >> ...
> >>>> Part of the problem with site-local was that the scope was ambiguous.
> >>> Agree.
> >>>> the term is not rooted in a discrete object with a position in the
> >>>> topology, contrast with autonomous system or prefix.
> >>> Just because of this point, it would better confirm the scope of
> >>> "site" when talking about it in case misunderstanding and confusion.
> >> It may be impossible. Actually I'd be very interested to hear any comments
> >> about the approach to defining address scope that we have taken in
> >> draft-carpenter-behave-referral-object. Maybe what we call a "limited scope"
> >> is a site? This should be discussed at a BOF in Hiroshima. Comments on the
> >> grobj mailing list please:
> >>  grobj at ietf.org
> >>  https://www.ietf.org/mailman/listinfo/grobj .
> >
> > I'm not sure it is impossible to define "site"; a site is
> > just a logical or physical partition (bounded by site border
> > routers) within a connected routing region. As long as the
> > nodes within the site remain associated with their site
> > border routers, they are still within the same site.
> >
> > Back to site-locals, my understanding was that RFC4193
> > ULAs were introduced in part to accommodate sites that
> > partition or merge. As long as each site border router
> > configures and advertises a its own ULA prefix, there
> > would be no ambiguity regarding the scope over which
> > the ULA applies.
> 
> That can change if you use the ULA over a VPN to another "site",
> or if the scope is chopped up with a NAT. I think it's rather
> a different thing for a human to understand what the effective
> scope is, compared with all hosts knowing algorithmically the
> scope of an arbitrary address. I think Dong Zhang is asking for
> an algorithmic definition.

I think what I was suggesting can be expressed algorithmically.
For a node within a site, as long as its site border routers
are still reachable and as long as its on-link or delegated
prefixes are still valid, then the node is still in the same
site. So, the site simply comprises the set of all nodes which
affiliate with the same set of site border routers in this way.

Additionally, the site is known by a name, e.g., "example.com".
Nodes within the site can resolve this name to determine the
list of site border routers.

Fred
fred.l.templin at boeing.com

>     Brian