> I think this is a terrible idea. I can envision many situations where a
> host would need both a site-local and a global address.

please, elaborate.  I haven't seen a good one yet.

> > In fact, I am suggesting that all routers and hosts treat
> > site-local unicast addresses exactly like global unicast
> > addresses, and that we administratively restrict the use
> > of site-local unicast addresses to non-globally-connected
> > networks.
> 
> I am comfortable with this. It's roughly what we do with RFC1918
> addresses today, and it has worked fine for many people. 

YES, this is what RFC 1918 _says_ - those addresses are for isolated
networks only.  And had those restrictions been adhered to I agree
it would have worked fine for most people.

NO, it's not roughly what we do with RFC 1918 addresses today - 
roughly what we do with them to day is use them as local addresses
behind a NAT - a usage which is NOT consistent with RFC 1918 and 
which has caused huge problems.

NO, this has not worked fine for many people - it's broken a 
large number of applications, and it has had a adverse effect on 
the network's ability to support apps.

> Link-locals are
> not routable, but site-locals are within the site, and since hopefully
> nobody will invent IPv6 NAT, filtering site-local addresses at the edge
> of the site seems the appropriate solution to me.

you're missing the burden that having a mixture of SLs and globals
puts on apps.

Keith
--------------------------------------------------------------------
IETF IPng Working Group Mailing List
IPng Home Page:                      http://playground.sun.com/ipng
FTP archive:                      ftp://playground.sun.com/pub/ipng
Direct all administrative requests to [EMAIL PROTECTED]
--------------------------------------------------------------------

Reply via email to