Paul A Vixie <paul@vix.com> writes:

makes one long for the days of HOSTS.TXT. which was at least reliable and coherent. maybe it'll make a comeback in the form of everybody-run-their-own "." server.

Hey, I take it back, a PVM-must-die thread on NANOG hasn't happened in a long time. Maybe NANOG isn't _utterly_ useless (only mostly)...

i guess i've said this about 900 times now. DNS is a coherent, distributed database. don't do anything that will make it less coherent.

Kill it outright. Build something that doesn't use the unreliable datagram protocol with all of its wonderful deficiencies, perhaps leaving the description of the database itself to the implementors (concentrating on protocol instead), and ideally avoiding the in-addr.arpa botch and the difficulties of scaling really large zones. Denninger and company are wonderful because they are forcing some issues which go right back to the days of HOSTS.TXT and the ARPA zone. I seriously hope they are able to forestall the insufficient bandaid approaches being proposed by more moderate technical people.

start back in on things that will make the same name mean something different (or become meaningless) depending on where you're standing when you resolve it.

But this is a really neat idea; have the DNS or its follow-on(s) represent _services_ and the utility of having this feature is fairly obvious. This is to some extent how NATs work, after all. A complaint about how '<port>.<zone>.<zone>...' can lead to confusion is a complaint about the lack of directory services through one can find a service in the first place. Sean.