Re: censored? corrected [Steve Pizzo cited in The Spotlight]
Originally DNS was just a handy user-friendly thing, but then it because the primary way to name a host. This allowed the IP address to become irrelevant, which allows entire sites to be renumbered when the network topology changes. This is important to keep the routing tables small (by minimizing the backwards-compatibility exceptions), which is a critical problem today. Thus all recent architectural decisions have been away from direct use of IP addresses anywhere.
Hunh? The Arpanet always used hostnames. Just about the only time you needed to specify a raw network address was when you were dialed into a TAC and you did things like "@o 137" to connect to mit-mc. DNS was created because the hosts.txt file, a single network-wide file that listed the name, address, aliases, system type, etc., of every host in the Arpanet got to be unmaintainable. Routing is totally unrelated to DNS, based as it is in IP addresses. Routers know nothing about DNS. /r$
In article <9511070647.AA00471@sulphur.osf.org>, Rich Salz <rsalz@osf.org> wrote:
Originally DNS was just a handy user-friendly thing, but then [...] Hunh?
The Arpanet always used hostnames.
Sorry. But still, then intent was to be user-friendly (right?), and a side effect was to make it possible to renumber without anyone noticing. Renumbering didn't affect the health of the net until recently, with CIDR, where blocks of IP addresses could be aggregated arbitrarily. A provider might now get 10.11.12/22 (i.e., a 22-bit-long prefix), and then assign customers 10.11.12/24, 10.11.13/23, &c., with only one entry in the backbone routing tables. To make this work you have to renumber when you change providers. Fortunately, we have DNS to provide a name more stable than the IP address. That's the connection between DNS and routing, and it's why using names instead of numbers is Good. -- Shields.
participants (2)
-
Rich Salz -
shields@tembel.org