[ih] early DNS, A paper that has something to to with the Internet

[Jack Haverty]

While I'm thinking of it, there's another part of Internet History and 
DNS...

IIRC, DNS appeared as a replacement for the files of "hostnames" that 
the NIC maintained but which had become unwieldy as the network grew.   
I don't recall the details, but it always seemed to me that the "Mail 
Records" aspect of DNS was sort of glued on to DNS as a convenient way 
to add another kind of data to the network service.    So DNS didn't 
just tell you what address to use to connect to a host, but also 
specified where to send its mail.

Then in the early 1990s, I was working at Oracle, and we had a similar 
need.   Database users (basically the hordes of people in corporations 
with some kind of terminal in front of them) wanted to connect to 
databases, by name.   E.g., some might connect to "Accounting", others 
to "Inventory", others to "Sales", "Personnel", "Physical Plant" etc.

At the time that struck me as a problem simlar to the DNS' Mail 
functionality.  One design choice could have been to add another record 
type or two to DNS, or perhaps even generalize DNS to include "Service" 
records, permitting numerous different services to then be referenced by 
name as Mail was.   Want to know where to send mail -- get the MX 
record.   Want to use Accounting -- get the SVC record.

That would certainly have been possible, but we didn't pursue it. It 
wasn't reasonable to rely on an Internet-specific mechanism because, at 
that point in time, essentially all corporations were using something 
other than TCP.  Although they usually had TCP somewhere in the 
organization (brought in by those radical new college grads no doubt!), 
most core functions of the company still ran on something else - SNA, 
DECNet, Netware, even Appletalk (seemed to be popular in Marketing).

We needed Naming that would work anywhere, even if there was no 
IP-capable network.  So we created some software called Oracle Names, 
which did pretty much the same things as DNS but would do it in any 
flavor of network and even across different networks by means of an 
"Interchange" (which was analogous to an IP gateway in the cloud 
diagrams I drew to explain it).

It turned out that was amazingly easy to do.   Once I was indoctrinated 
into the world of databases, by the immersion osmosis method of being 
surrounded by people who all lived and breathed databases, it was easy 
to see DNS, and naming mechanisms in general, as a fairly simple and 
straightforward database  program, built on top of the distributed, 
replicated, database technology that we could assume was going to be 
present in our customers' networks.

Timing seems to be an important factor in the History of the Internet.   
Databases (relational) weren't really prominent, especially in the 
academic and research communities I experienced, in the early 80s, and 
SQL, i.e., the protocol which makes databases easy to use, hadn't yet 
fully congealed as a standard.   I had heard of databases, but knew 
little about them before joining Oracle.

I wonder how the Internet mechanisms would have evolved if the timing 
had been a bit different, and the people who built the early network 
mechanisms had been able to consider using database technology or 
techniques when designing  mechanisms such as DNS.

/Jack Haverty



On 7/20/21 11:43 AM, John Levine via Internet-history wrote:
> It appears that Stephane Bortzmeyer via Internet-history <bortzmeyer at nic.fr> said:
>> This is not a criticism: his goal was not to do a survey of name
>> resolution systems, but to produce something which worked (which he
>> did). When designing a new system, there is a fine line to draw
>> between "studying so many variants you never have time to actually
>> create something" and "jumping on the first idea without considering
>> the other possibilities". IMHO, the choices were basically the correct
>> ones, giving the techniques of the time (blockchain was not invented yet :-)
> If you look at the design criteria for the DNS, good read-only performance via replication, highly
> scalable, distributed management, there weren't a lot of plausible design options and there still aren't.
>
> In fairness, when people were designing the DNS in the early 1980s I doubt it occurred to anyone
> that less than 20 years later, DNS names would have become collectable and tradable fashion accessories.
>
> R's,
> John