[ih] UDP Length Field?

[Jack Haverty]

At the time (late 70s), there was another issue related to UDP and
real-time applications.   The Internet was in its "fuzzy peach" stage of
development, where almost all long-haul links were established by
interconnecting gateways at user sites to the ARPANET.  

The ARPANET was effectively a "byte stream" service, where everything
sent in from a host was delivered to the host at the other end intact,
in order, and reliably.  This made it difficult to imagine much useful
experimentation with UDP and lossy transmission, since the ARPANET would
never lose anything, and retransmit internally as needed to deliver the
data in order.

There was an "uncontrolled mode" of operation possible within the
ARPANET, which wouldallow a user computer to send packets bypassing all
of the reliability mechanisms.   That could be used to deliver datagram
Internet service.   However, the ARPANET managers (at DCA and BBN) were
extremely reluctant to permit hosts (e.g., gateways) to use that mode,
for fear that the uncontrolled traffic would crash the ARPANET.   I
recall being involved in several "discussions" about using ARPANET
uncontrolled mode for Internet experiments, but I don't remember any
permissions ever being granted.   John Kristoff's mention of the
problems today's operators are facing was a reminder that the issue of
how to control datagram service still has not been solved -- other than
by turning it off.  

UDP functionality is about more than defining a protocol....

/Jack Haverty


On 12/2/20 12:13 PM, Vint Cerf via Internet-history wrote:
> David Reed's timing comports with what I remember. I got to ARPA in Sept
> 1976. TCP2 came out in (early?) 1977, then the split of TCP/IP; TCP 2.5,
> TCP 3 (at which point I think the IP doc and new TCP doc appear. Then
> TCP/IP v4 in 1978 (from memory; have not checked RFCs). David is also right
> about Internet and TCP going along in parallel with ARPANET and documents:
> Internet Experiment Notes. We did publish RFC 675 in Dec 1974 Internet
> Transmission Control Protocol.
>
> I have a fairly clear recollection of being at USC/ISI with Jon, Danny
> Cohen and David Reed in 1977 discussing the real-time application needs
> that TCP would not satisfy owing to retransmission delays.
>
> v
>
>
>
>
> On Wed, Dec 2, 2020 at 1:07 PM David P. Reed <dpreed at deepplum.com> wrote:
>
>> Noel - you are plain wrong that UDP happened after the split.  UDP was
>> created in 1977, by me and Jon Postel and Danny Cohen, primarily. The
>> "spec" was a sketch in Jon's notes, not an RFC. (RFC's dont tell the full
>> history, not by far! Especially back then, where the TCP working group was
>> working on its own, separate from the ARPANET process of RFC's. Our work
>> was in email and in meetings. The split was done on the blackboard at
>> Marina del Rey, with IP, TCP and UDP all defined. UDP at that time
>> satisfied Danny, me, and John Schoch who were the main "users" demanding a
>> datagram user level.
>>
>>
>>
>> As I recall, you, Noel, were not involved at all in the Internet project
>> at that time (nor was Dave Clark directly). I was the guy. I had spent
>> summer 1976 designing DSP, and in the fall Bob Kahn and Vint strongly
>> discouraged continuing with DSP for LANs, and encouraged me to join the TCP
>> project to bring my ideas into that framework. Which I did, until early
>> 1978, when demands of completing my doctoral thesis forced me to stop
>> direct participation and Dave Clark became involved.
>>
>>
>>
>> As I recall, you were engaged with token ring hardware during that time,
>> not TCP or IP, along with Clark and Pogran, right?
>>
>>
>>
>> PS: I've given up, mostly, on trying to help clarify Internet "history".
>> Because of the egos involved seeking credit as the "father", claiming the
>> Internet was just ARPANET (BBN) and not a separable concept about
>> internetworking, etc. what I find is that books like Katie Hafner's BBN
>> propaganda are accepted as the truth, along with propaganda from UCLA etc.
>> In fact, the history is far more complex than these tales of "heroic"
>> inventions of things like the "@" that is said to be Tomlinson's only
>> contribution! (Ray did FAR more, including sorting out sequence numbering
>> and encouraging the use of 32-bit oriented frame structures, even on 36-bit
>> machines like PDP-10's and GE645's.
>>
>>
>>
>> I have very little interest in getting "credit" to stroke my ego, unlike
>> some in the community. But I do wish people would get it right.
>>
>>
>>
>>
>>
>> On Sunday, November 29, 2020 2:28pm, "Vint Cerf" <vint at google.com> said:
>>
>> Noel,
>> yes, we did the split to support real-time and then concluded that UDP was
>> the best way to present the "service" vs running over raw IP.
>> v
>>
>> On Sun, Nov 29, 2020 at 1:46 PM Noel Chiappa via Internet-history <
>> internet-history at elists.isoc.org> wrote:
>>
>>>     > From: Craig Partridge
>>>
>>>     > Recall that the creation of UDP meant TCP and IP had to be split
>>> apart
>>>
>>> No. The TCP/IP split _long_ predates the creation of UDP. The former is
>>> already apparent as of IEN-21, "TCP 3 Specification" (see pages 56 and 59
>>> for
>>> the IP and TCP header formats), from January 1978. UDP is IEN-71, from
>>> 21-Jan-79 (and as I recall, there was not a lengthy discussion before it
>>> came
>>> out).
>>>
>>> Oh, looking at IEN-71, in the packet format description, it says "data,
>>> padded
>>> with zero octets at the end to make a multiple of two octets". So Vint's
>>> comment
>>> about the length was right on target.
>>>
>>> It mentions host name lookup (_not_ DNS; it was servers which had a copy
>>> of
>>> the host table) as the intended appplication. Time was also early, IIRC.
>>> My
>>> recollection is that TFTP was the first non-datagra protocol (i.e. not
>>> single-packet transactions) to make use of UDP, but my memory mmay be
>>> failing
>>> me there.
>>>
>>>    Noel
>>>
>>> --
>>> Internet-history mailing list
>>> Internet-history at elists.isoc.org
>>> https://elists.isoc.org/mailman/listinfo/internet-history
>>
>> --
>> Please send any postal/overnight deliveries to:
>> Vint Cerf
>> 1435 Woodhurst Blvd
>> McLean, VA 22102
>> 703-448-0965 <(703)%20448-0965>
>> until further notice
>>
>
>