Re: Domain names for ISP infrastructure links
That may sound silly, but there's a merit in the idea that domain names can successfully replace the SNMP's object IDs. SNMP and DNS then become the same protocol. In fact, all protocols requiring exchange of some request and getting some reply from a uniquely named object can be made a particular case of a generic protocol. There's no reason for Internet to be a hodge-podge of specialized kludges. In most cases, it only takes an ability for DNS to "hand off" the continuation of name resolution to application programs. See -- i don't think everything Plan 9-ish is pure crap. --vadim From: John Curran <jcurran@bbnplanet.com> No... we're going to mandate OID-style router names for the world's Internet providers: <isoc-tree>.<internet-infrastructure branch>. <iso country code>.<city ordinal>.<provider #>. <facility #>.<rack #>.<shelf #>.<router ordinal> Great fun to watch (it could be happening this way in the alternate universe next door.) ;-) /John
On Tue, 7 Jan 1997, Vadim Antonov wrote:
That may sound silly, but there's a merit in the idea that domain names can successfully replace the SNMP's object IDs. SNMP and DNS then become the same protocol.
Cool. I can now dream of a day when nslookup for a router returns something lovely like: 1.3.6.1.2.1.11 -dorian ;)
That may sound silly, but there's a merit in the idea that domain names can successfully replace the SNMP's object IDs. SNMP and DNS then become the same protocol.
In fact, all protocols requiring exchange of some request and getting some reply from a uniquely named object can be made a particular case of a generic protocol. There's no reason for Internet to be a hodge-podge of specialized kludges. In most cases, it only takes an ability for DNS to "hand off" the continuation of name resolution to application programs.
This isn't that far fetched, actually; look at all the fun you can have with the Hesiod class, or look at how contact info is stored in DNS within the reverse address blocks for the MAEs. Even such wonderful things as password files can be distributed that way, or POP mailbox locations; for example, my SO works at an institution that distributes mail for 40,000+ users among 12 mail hubs by querying TXT records out of DNS. It's fast, it caches well, it scales beautifully, and best of all, it didn't require a whole new protocol! In fact, the password file entries for all 40,000 users are also retrievable via simple DNS queries, so that building up customized /etc/passwd files on remote machines is a simple task. And no, they're not silly enough to put the encrypted password in the DNS entry; since they use Kerberos, that field is left with an invalid string (they could even have put fake encrypted entries, just to torment would-be hackers), and the actual authentication is done via encrypted Kerberos calls. The mailbox lookups, and alias expansions, and forwarding info is done via hesinfo lookups that are standard Sendmail 8.8.x, no local hacks needed. The only step that would make things even cooler would be having the getpw* calls modified in the system libraries to do the DNS queries directly to get user id's, home directories, and the like, rather than having to dump them periodically to the flat /etc/passwd files. Sorry, didn't mean to rant so much; it's just that when you get right down to it, the DNS hierarchical naming system can be used to store so many other forms of information beyond just the plain and boring hostnames that most of us use it for right now. It's a shame it's so underutilized. Time to hop off my soapbox now... Matt Petach
See -- i don't think everything Plan 9-ish is pure crap.
--vadim
From: John Curran <jcurran@bbnplanet.com>
No... we're going to mandate OID-style router names for the world's Internet providers:
<isoc-tree>.<internet-infrastructure branch>. <iso country code>.<city ordinal>.<provider #>. <facility #>.<rack #>.<shelf #>.<router ordinal>
Great fun to watch (it could be happening this way in the alternate universe next door.)
;-) /John
participants (3)
-
Dorian R. Kim
-
Matthew Petach
-
Vadim Antonov