On Tuesday 13 January 2009 06:27:10 Hui Deng sent:> May I chime in, I feel identity type is a good idea.> > But if you map DNS to other identity,> how network socket connection could work in that case/ Shortly, like socket connection for anycast TCP does.You tune right in, see identity type is useful namely for connection inititation and re-establishment. It thereforе shall natively underlie IP mobility. > > thanks for your consideration> > -Hui> > 2009/1/12 Toni Stoev <ietf@xxxxxxxxxxxxxx>> > > On Monday 12 January 2009 00:51:24 TSG sent:> > > Toni Stoev wrote:> > > > Hi,> > > >> > > > DNS job> > > >> > > > When a connection to a network node is to be initiated its DNS name is> > resolved to an IP address which shows the location of the node on the> > network. So network nodes are findable by name even if their locations> > change.> > > >> > > I think you are backwards... The nodes are still reachable if they> > > change their physical location or the assigned networks addresses> > > temporarily mapped too those names by DNS or DHCP.> >> > Findable, not reachable, even just identifiable, is what is essential for> > the naming system DNS.> >> > > > The job of the DNS is to identify a node by its name. Another job is to> > maintain information about node's current location so a new connection can> > be started at any time.> > > >> > > Hmm... I think the job of DNS is to map the 'text based system names' to> > > a network address that the routing infrastructure and features can> > > create a set of pathways for that data to reach said system.> >> > The job of the Domain Name System, not routing or delivery system, is to> > map names to identity. Furthermore, identity mapped to may be of entire> > routing domains, aggregating nodes.> >> > > > Why should DNS be bothered with connectivity and/or topology status?> > > > There must be a distinct mechanism to handle mapping of node identity> > to network location.> > > >> > > GeoPriv would have you believe that another> >> > Still network location is disticnt from geographic.> >> > I suggest that there be a networking asset, an address type, that> > represents network identity of nodes. And addresses of that type be resolved> > to by DNS name queries and be mapping to unicast network addresses.> > This way DNS shall be unburdened of connectivity, routing and reachability> > issues related to named nodes.> > _______________________________________________> > Ietf mailing list> > Ietf@xxxxxxxx> > https://www.ietf.org/mailman/listinfo/ietf> >> _______________________________________________Ietf mailing listIetf@xxxxxxxxxxxxx://www.ietf.org/mailman/listinfo/ietf