RE: I-D ACTION:draft-ietf-shim6-applicability-01.txt

[Sam Xia <xiazhongqi@huawei.com>]

Hi,
  Even though shim6 protocol is host-based protocol, shim6 provides the
mutihoming capablity to the hosts in the multihoming site. Furthermore, the
multihoming model for shim6 is a site model, not host model. So I think that
this is the reason that the authors of shim6 call shim6 as a site solution.
  But I think shim6 maybe be also useful in the host-multihoming context.
Why do the authors not mention the host multihoming at all?

 Best Regards, Sam Xia 

> -----Original Message-----
> From: owner-shim6@psg.com [mailto:owner-shim6@psg.com] On 
> Behalf Of Iljitsch van Beijnum
> Sent: Tuesday, June 13, 2006 10:01 PM
> To: Daniel Roesen
> Cc: shim6@psg.com
> Subject: Re: I-D ACTION:draft-ietf-shim6-applicability-01.txt
> 
> On 13-jun-2006, at 14:35, Daniel Roesen wrote:
> 
> >>> Given that shim6 doesn't provide site multihoming, but only host 
> >>> multihoming, I'd suggest to change this wording to 
> reflect reality.
> 
> >> The shim6 approach is to provide site multihoming by adding a 
> >> capability to multi-home to the hosts within the site, as you know.
> >> Do you think that is semantically misleading?
> 
> > Yes I do. It's not a site being multihomed, it's a host being 
> > multihomed. Which is (to me and many others) a totally different 
> > story.
> 
> Is it hair splitting day already?
> 
> > There is absolutely nothing in shim6 which leans on the "site"
> > concept, no site-wide policy enforcement etc. A site multihoming 
> > solution involves a site-wide routing policy (and it's 
> enforcement), 
> > totally independent of the hosts which are usually to be considered 
> > untrusted and under different administration than the site network.
> 
> > shim6 doesn't deliver anything of that, so it's not a site 
> multihoming 
> > solution. It's a solution for ad-hoc opportunistic host 
> multihoming in 
> > cases where both ends support it and where two or more 
> transitted IPv6 
> > addresses with associated uplinks are available.
> 
> > Unless I've misunderstood all of that. :-)
> 
> My understanding of the whole thing is that either you have 
> two or more cables coming into your SITE or you have two or 
> more cables coming into your HOST. The former should be 
> called site multihoming, the latter host multihoming. 
> (Allowing for wireless multihoming while still making the 
> point as intended is left as an exercise for the
> reader.)
> 
> I don't think that under the above definition shim6 can be 
> considered a mechanism for host multihoming, even though it's 
> not impossible to use it as such. (But then I can run BGP on 
> my laptop and do "host multihoming" in that way too.)
>