[ih] DNS origins?

[Jack Haverty]

It's been a long time....   I remember the debate, and that it was 
somewhere in California.  I thought it was San Diego, but might have 
been PARC.  There was a large auditorium with theater-style seating 
where the discussion occurred.

This meeting may have happened before the D appeared in DNS.  There had 
been ongoing discussions of how to do a Name Server for the Internet, 
since the ARPANET style mechanism using a file at SRI-NIC couldn't 
handle the obvious impending torrent of workstations on LANs that TCP 
made possible for the network community.   It was one of the topics on 
the ICCB's list of "things we have to work on". Jon had published 
several IENs on "Internet Name Server", starting IIRC with #61.

The discussion I remember concerned the methods and timing involved in 
updating a name/address pair when there was a collection of servers on 
the Internet.   The "simple" mechanism could take as much as a day or so 
for any change to propagate out to reach all the servers.  That was of 
course incompatible with the changes expected with mobile hosts, or even 
LANs.   It was obvious even then that it was much easier to move a 
Sparcstation than a PDP-10.   I don't recall anyone ever mentioning 
smartphones or Dick Tracy watches as perhaps the ultimate "mobile 
host".   But we have them today and the Internet still seems to work...

The early versions of the protocol, again IIRC, structured names 
somewhat like IP addresses, with "Net" and "Rest" parts.   But that 
would have problems if a computer moved from one network to another (no 
one had laptops yet...and no wifi either)

So, at some point that "Internet Name Server" approach was changed to 
incorporate Domains so that names didn't necessarily correspond to any 
physical network structure, and anybody could pick whatever names they 
wanted in their own world/domain.

Roughly at about the same time, Jon Schoch wrote his seminal "Names, 
Addresses, and Routes" paper, which brought a lot of thinking into a 
cohesive idea.   Xerox protocols (PUP et al) had their own schemes for 
naming, and PARC people did participate in Internet meetings. History 
IMHO should look at what was going on in other communities at the same 
time.  The Internet did not evolve in a vacuum.

I remember all this stuff happening (and namedroppers, another of those 
mailing lists where lots of technical interaction happened). But dates - 
not so much...   So, depending on what the "origins of DNS" means, I'd 
point to Jon's earliest IEN as the start of Name Servers on the Internet.

/Jack


On 6/9/21 2:26 PM, Barbara Denny via Internet-history wrote:
>   I am curious about the dates you suggest.   I was in the mobility perspective camp but I don't remember anyone telling me that this had been discussed before or providing me any  information in this context at the time.  I didn't start working on packet radio until the end of 1981, or perhaps the beginning of 1982, so anything done  prior to those dates I might not have been aware of.  However, I would have thought other people at BBN would have mentioned it since I am  pretty sure Jil Westcott, BBN packet radio project manager,    is the one who gave me the assignment for the meeting.
> barbara
>
>      On Wednesday, June 9, 2021, 01:16:37 PM PDT, Jack Haverty via Internet-history <internet-history at elists.isoc.org> wrote:
>   
>   I remember a lengthy discussion and somewhat heated debate about the
> design for the Internet's name system at the Internet Meeting held in
> San Diego, IIRC at Linkabit.  Sorry, I can't remember exactly when that
> occurred, but it was one of the "winter" meetings which were always held
> somewhere in California.   Pretty sure it was before 1982, probably 1980/81.
>
> There were two "camps" involved in the debate.  One was arguing for
> powerful mechanisms to handle updates of name/address mappings; the
> other was arguing for more simplicity.
>
> I remember asking the two camps to explain what problem they were trying
> to solve.   One camp was focussed on ARPANET-style host computers, which
> changed their IMP ports very rarely.   Expectations were that Internet
> addresses would change in a similar pattern. The other camp was focussed
> on what could be called the "mobile host" problem, exemplified by the
> various Packet Radio experiments that had been going on.   Their
> expectation was that IP addresses might change rapidly and frequently,
> in the heat of a battlefield operation.
>
> These were obviously very different problems, motivating very different
> solutions.  IIRC, the debate led to the DNS implementations and specs
> not long after that meeting in San Diego.
>
> Note that the notion of "Internet Name Server" existed before DNS - see
> IEN 89 -- https://www.rfc-editor.org/ien/ien89.txt  and 116 -
> https://www.rfc-editor.org/ien/ien116.txt
>
> I'll have to look through my old notebooks from the 80s...
>
> /Jack Haverty
>
>
> On 6/9/21 12:29 PM, Dave Crocker via Internet-history wrote:
>> On 6/9/2021 11:17 AM, Barbara Denny via Internet-history wrote:
>>>    I remember thoughts about DNS were developed enough by summer of
>>> 1983 that  I was asked to prepare a talk about DNS and packet radio
>>> at what I believe was the last packet radio meeting.  Unfortunately I
>>> don't remember what I used to learn about DNS so I could prepare my
>>> thoughts. I remember this is where I met Jon Postel but I don't
>>> remember if Paul Mockepetris was there.
>>
>> Some additions about timeline:
>>
>>
>> I had nothing to do with the creation of any aspect of the DNS.
>>
>>
>> However RFC 822, defining Internet mail format -- with relatively
>> small modifications from RFC 733 --as published August 1982. It
>> included support for domain name, which is to say support for the
>> dotted name notation in a host reference.
>>
>> SMTP also added domain name support, at the same time. (duh. Written
>> by Jon.)
>>
>> I do not remember the details of how the directive to add this support
>> in RFC 822 developed nor how I was told of the syntax. 822 was
>> developed through group discussion, over email.  I don't even recall a
>> face-to-face meeting for it.  SMTP definitely did have f2f sessions.
>>
>> I only recall one discussion with Jon, concerning the handling of
>> domain names in SMTP, where I was confused that it always passed the
>> entire domain name, rather than stripping off the right-hand field, as
>> the message transited a hop.  I had not yet understood that this was
>> not a source route.
>>
>> So I believe the general concept of the administrative/semantic
>> hierarchy -- distinct from the distributed operational query mechanism
>> -- was fully set by Fall of 1982.  (I'm not saying the latter wasn't
>> but that I don't know anything about that part of the design timeline.)
>>
>> d/
>>
>