The 7 Layer model is a useful tool to talk about things, its not a rei-fied thing. That said, apparent layer violations invite critique because they inherently carry architectural consequence.
I think the overloading of a (semantic space) name to have special properties to take it out of the system is a case in point. Its an application through session layer property: packets don't get sent using .onion labels in source/destination fields. And the imposition on all the other layers to handle .onion specially, feels (to me) a mortal wound. This, compared to the cost of taking syntactic limits in the URI, and applying a lever there to wedge :tor: into the URI form, denoting what you want to happen. SOCKS is pretty much a shim. Its a clean layer impact. Its (to me) like taking fopen() and replacing it by bzip2fopen() to get silent bz2 capability in existing file I/O On Thu, Dec 10, 2015 at 1:02 PM, John R Levine <jo...@taugh.com> wrote: > With onion you get a rather different thing that looks like an open >>> TCP connection, a couple of levels up the protocol stack. >>> >> > Strictly an Onion address yields you a _real_ TCP connection to your SOCKS >> server, ... >> > > It's certainly a virtual circuit, but it's not a TCP connection because > the endpoints aren't IP addresses. > > The Onion addresses aren't making a "protocol switch", ... >> > > Really, they are. You can't do a DNS lookup on the address, there is no A > or AAAA record with an IP address to which you can open a TCP connection. > > Regards, > John Levine, jo...@taugh.com, Taughannock Networks, Trumansburg NY > Please consider the environment before reading this e-mail. > > > _______________________________________________ > DNSOP mailing list > DNSOP@ietf.org > https://www.ietf.org/mailman/listinfo/dnsop >
_______________________________________________ DNSOP mailing list DNSOP@ietf.org https://www.ietf.org/mailman/listinfo/dnsop
